Docu45799 Celerra Network Server Release Notes 6.0.70.4

1

EMC® Celerra® Network Server

Version 6.0.70.4

Release Notes P/N 300-009-958

REV 18 February 28, 2013

These release notes contain information on these topics:

Product description ................................................................................... 2 New features and changes ........................................................................ 2 Known problems and limitations .......................................................... 27 Problem prevention ................................................................................. 50 Technical notes ......................................................................................... 56 Environment and system requirements................................................ 78 Documentation ......................................................................................... 79 Installation and upgrades ....................................................................... 81 Fixed problems ......................................................................................... 88 Version 6.0.70.4 ......................................................................................... 89 Version 6.0.65.2 ......................................................................................... 98 Version 6.0.60.2 ....................................................................................... 115 Version 6.0.55.3 ....................................................................................... 129 Version 6.0.51.6 ....................................................................................... 141 Version 6.0.43.1 ....................................................................................... 189 Version 6.0.43.0 ....................................................................................... 189 Version 6.0.42.3 ....................................................................................... 190 Version 6.0.41.4 ....................................................................................... 222 Version 6.0.41.3 ....................................................................................... 225 Version 6.0.40.8 ....................................................................................... 301 Troubleshooting and getting help ....................................................... 326

2

Product description

EMC Celerra Network Server Version 6.0.70.4 Release Notes

Product description

The EMC Celerra Network Server is a network-attached storage (NAS) server. It enables clients on a network to store and retrieve files over the network. These release notes contain information about version 6.0 of Celerra Network Server.

Check EMC Online Support at http://Support.EMC.com for updates to this document. Updates are made when a new version or patch is released, or when new information is discovered.

New features and changes

Version 6.0.70.4

NS gateway support for new Symmetrix® Enginuity™ 5876.159.102.

• NS gateway supports Enginuity 5876.159.102 running on any of the VMAX platforms (10K, 20K, and 40K).

• NS gateway software features such as TimeFinder® and SRDF®, which are supported on the earlier Symmetrix Enginuity, will be supported on 5876.159.102 Enginuity.

• Software features such as FAST™ and Federated Tiered Storage (FTS) are not supported with 6.0.70.4.

• Thin BCV (Business Continuance Volume) disks are supported on NS gateway. • NS gateway inherits all hardware and software restrictions imposed by VMAX

10K, 20K and 40K, and Enginuity 5876.159.102. • VP Compression is supported on NS gateway. It is recommended that you

create a user defined pool for the compressed LUNs so that consistent performance can be provided for the file systems. VP compression must be done from the Solution Enabler host.

• I/O Quota is supported on NS gateway. It is recommended to create a user-defined pool for the LUNs that have the same I/O Quota configuration so that the user-defined pool can provide consistent performance for file systems. I/O Quota must be configured from the Solutions Enabler host.

• The recommended configuration (best practice) for NS gateway using I/O Quota with masking view (LUN masking feature) is as follows:

Ensure that a parent Storage Group is in a masking view with the Data Mover initiator group and Symmetrix FA port group.

http://support.emc.com/

3



Create multiple Child Storage Groups with different performance expectations. For example, extra performance, performance, and capacity.

Set an appropriate I/O Quota in each Storage Group. Execute diskmark operation. If the storage groups are all from fibre channel, create user

defined storage pools for each storage group. Create file systems from the user defined storage pools.

Version 6.0.65.2

For NS-G2, NS-G8, VG2, VG8 gateways, if you plan to upgrade to VNX File OE 7.0 or later in future, do not install this version.

Versions 6.0.61.0, 6.0.60.2, 6.0.51.6, 6.0.43.1, 6.0.43.0, 6.0.42.3, and 6.0.41.4

There were no new features or changes to functionality in these versions. The Fixed problems section contains changes made in this version.

Version 6.0.55.3

NS-G2/NS-G8 6.x support for VNX Block arrays

This release supports installation of NS-G2/NS-G8 Gateways and their connection with VNX Block arrays using the Fibre Channel protocol. New features in the upcoming VNX Operating Environment for Block 5.32 are not supported.

Version 6.0.41.3

Checkpoint scheduling relative name convention

As of this release, SnapSure supports checkpoint CVFS names that are dynamically assigned based on the relative order of checkpoints (for example, nightly.001, nightly.002, nightly.003, and so on.) To use this option, you must specify the relative CVFS name prefix, the delimiter, and the starting index. The delimiter and starting index values are recorded in the nas_param database and must be manually edited. The complete CVFS name while using the relative prefix option is a concatenation of the prefix, delimiter and starting index.

The prefix of a relative checkpoint must be a PFS-wide unique string and can contain up to 20 ASCII characters. It must not include intervening spaces, colons (:), or slashes (/). The delimiter must be either a period (.)

4



or an underscore (_). The starting index must be either 0 or 1.

This feature is managed through the GUI or the CLI, or both.

CPU Utilization Threshold Alerting

As of this release, you can enable CPU utilization threshold alerting and receive automated SNMP trap or email notifications when a Data Mover CPU utilization crosses a certain threshold for a set period of minutes and persistence. You can also be notified when the condition is no longer applicable and the CPU utilization subsides to a normal level.

The Configuring Celerra Events and Notifications technical module provides more detailed information about this feature.

File system replication with common base checkpoints

This feature is also known as "Incremental Attach Replication".

When you replace source or destination file systems, the replications and internal checkpoints are deleted. You will have to perform a full data copy to be able to create future replications with the original configurations. To avoid performing a time-consuming full data copy over the WAN, you can create a local replica of the file system that you want to replace.

You can then create one or more user-specified checkpoints on the file system to be replaced, the local replica, and each of the other file systems involved in the replication.

After creating user-specified checkpoints, you must refresh each replication by specifying the user checkpoint on the source and destination file system for that replication.

Consequently, the user checkpoint of each destination file system (including local replica) has the same view of the file system as the user checkpoint on the source file system that will be replaced. User checkpoints on the source and destination file systems, which have the same content, are called common base checkpoints. You can now use the local replica for replications without performing a full data copy over the WAN.

Implementing this technique is useful in disaster recovery scenarios as well. To avoid a full data copy in the event of a disaster, create one or more user-specified checkpoints on the source and destination file systems and refresh their replication by using these user checkpoints. This process transfers data from the source user checkpoint to the destination file system through the existing replication session. Thus, the destination user checkpoint has the same view of the file system as the

5



source user checkpoint. This process is repeated for each destination in the one-to-many configuration. For a cascading replication, the replication must be refreshed for this first hop and then for the second hop.

After this process, even if you delete the replication sessions, you can configure new replication sessions between any two destinations because replication automatically identifies the common base checkpoints and avoid full data copy. You can create user checkpoints for file system and VDM replications. For VDM replications, you can create the user checkpoints on the root file system of the VDM. You cannot delete user checkpoints when they are being used by replication either for transfer or as an active common base.

Greater than 4 TB Quotas

As of this release, the storage limit per user, group, or tree quota has been increased from 4 TB to 256 TB. You can control the usage of the disk resource and limit the expensive disk space usage by various user or group or trees on a file system. Support for 256 TB quotas does not imply any change in the maximum size of the underlying file system.

Caution: If you replicate a quota-enabled file system from a system running Celerra Network Server (CNS) version 6.0.41.0 or later to a destination system running a CNS version earlier than 6.0.41.0, the destination file system can become unusable.

Upgrade quota database limits for file systems You can upgrade your quota databases to the new 256 TB limit from the existing 4 TB limit. The nas_quotas command has a new option called -quotadb that, when run, performs the upgrade. The file system will be unavailable during the upgrade.

Note: Before the upgrade process runs, the Control Station displays the estimated upgrade time on the file system whose quota database will be upgraded, and also displays a warning message to notify you that the file system will be unavailable during the upgrade process. If you are in interactive mode, a dialog displays letting you choose whether you want to continue. Since the file system will be unavailable during the upgrade, the upgrade can be deferred to a later time. If you are in non-interactive mode, after displaying the estimated upgrade time message and warning message, the upgrade process starts immediately.

The Command Reference Manual contains more detailed information about the -quotadb command.

6



Online quota update tool

Once a tree, user, or group quota is enabled on a directory, in some instances, differences may occur between a reported quota usage and the actual usage. The online quota update tool lets you run a quota check on a tree or file system. The tool updates the tree, group, and user usage statistics, while keeping the directory tree and file system online, and keeping data access available. You can also run a quota check in the offline mode, if desired.

If you have file system administrative privileges, you can use the Celerra CLI to run the online quota update tool. If an error is encountered, an error message is generated. The status of the online quota check shows the percentage completed.

Note: 512 MB of free space is needed on the file system to run a tree quota check.

Performance problem collection tool

This tool enables users to send all relevant information related to the file system to EMC for diagnosis. The performance problem collection tool is a single CLI tool that gathers all the relevant logs (systat, IO trace, etc.) for a specified file system so that it can be easily transmitted to EMC as part of a support escalation. Prior to this release, users had to manually provide this information.

Symmetrix Enginuity version 5875

This release supports the new Symmetrix Enginuity version 5875, which runs on the VMAX platform, and includes a new version of Solutions Enabler v7.2. VNX supports tolerance for Enginuity 5875 release only. It does not leverage any new features included in 5875.

User-defined pool usage enhanced

The -is_greedy attribute of the nas_pool command has been enhanced to include identifying which member volumes of a user-defined storage pool are used to provide space when creating or extending a file system.

If the -is_greedy attribute is set to n (default), the system uses space from the user-defined storage pool's existing member volumes in the order that the volumes were added to the pool to create a new file system or extend an existing file system.

If the -is_greedy attribute is set to y, the system uses space from the least-used member volume in the user-defined storage pool to create a new file system. When there is more than one least-used member

7



volume available, AVM selects the member volume that contains the most disk volumes. For example, if one member volume contains four disk volumes and another member volume contains eight disk volumes, AVM selects the one with eight disk volumes. If there are two or more member volumes that have the same number of disk volumes, AVM selects the one with the lowest ID.

When extending a file system, the -is_greedy attribute is ignored unless there is not enough free space on the existing volumes that the file system is using.

Windows event log auto archive

With Windows operating system, applications can use the event logging mechanism to log their own events. Celerra currently supports three such event logs that are security, system, and applications.

The physical format of these logs use a Microsoft format called 'evt' that has a limitation of 4 GB in size because there are some fields stored on 32 bits integers. Windows 2008 has introduced a new format 'evtx' that does not have this limitation.

The event log auto archive feature allows you to automatically archive an event log on a particular trigger and to continue the logging on a new event log without losing any events. The archive is triggered on a time and/or on an event log size basis defined by parameters in the Windows registry.

You can specify a retention policy to keep the event log archives before they can be recycled based on the duration and/or the total archive disk size.

All parameters are stored in the Windows registry of each VDM. Therefore, each VDM will have its own configuration. The parameters can be viewed and edited with tools like regedit. The Configuring and Managing CIFS on Celerra technical module provides more information about this feature.

Version 6.0.40.8

Celerra Upgrade Tool -localization upgrade support

The Celerra Upgrade Tool (CUT) allows you to download language packages and manage your Celerra in your primary language. The Download Software wizard helps you download the language packages from the EMC Online Support website based on your Celerra configuration. The Celerra Upgrade wizard then installs the language

8



packages on your Celerra. Celerra file services will not be disrupted during language package installation. This feature is supported on EMC Celerra 6.0 systems only.

Unisphere upgrade procedure

EMC Unisphere Service Manager (USM) is a collection of tools that help you update, install, and maintain your CLARiiON system. It also includes the Celerra Upgrade Tool (CUT) which consists of tools that help you perform software upgrades to Celerra. This procedure explains how to use USM to upgrade Celerra Unified Platform systems.

1. Go to EMC Online Support and select Support by Product > Celerra NS> Downloads.

2. Select the USM installation (executable) file (for example, Unisphere Service Manager (USM) (Windows) 1.2.2.1.0020) and download the file to a known location on your PC.

3. In the folder where you save the USM installation file (for example, UnisphereServiceManager-Win-32-x86-en_US-1.0.0.1.0490-1.exe), double-click the executable (.exe) file, and if necessary, click Run to start the installation wizard.

4. Follow the instructions that appear. After you accept the license agreement, the wizard verifies that the server is running the supported Java Runtime Environment (JRE) version.

If the server is running an earlier version of the JRE, select Yes when asked if you want to continue with the installation.

The installation wizard prompts you to download and install the later version of the JRE from the Sun website.

If the server is running a later version of the JRE, the installation wizard asks if you want to continue with the installation.

Select Yes to continue or No to quit the installation.

If you select Yes, the USM may not work properly with a later JRE version. If you select No, and manually remove the later version of the JRE, other applications that require this version may not work properly.

5. When the installation is complete, click Done.

http://java.sun.com/

9



Register your Celerra to use the Celerra Upgrade Tool (CUT) 1. Register your system on EMC Online Support to enable CUT to

download software. Go to the Service Center tab and select Register a Product.

2. Enter the Product ID and click Continue.

3. You are now entitled to use CUT.

Run the Unisphere Service Manager The USM will allow you to upgrade the NAS and FLARE software on your system. As a general rule, you should update the NAS version to the latest qualified release (6.0) prior to updating the FLARE version to its latest release (R30).

Upgrade NAS 1. Double click the Unisphere Service Manager icon on your desktop.

2. Click Login and log in to a Celerra. CUT supports version 5.6 system login as root only. The Name field is prefilled with the username root. The root password is required.

10



3. In the menu bar, click Software > System Software to run the Celerra Upgrade Tool. Three wizards will guide you, step-by-step, through the process of upgrading the Celerra software, which includes updating the Control Station and the blades.

4. Select the wizards in the order presented on the screen:

Download Software Pre-upgrade Health Check Perform Software Update

The Celerra Upgrade Tool online help provides information about using the wizards.

5. Select the CUT files for download. For example:

6.0.36.4_apps.iso 6.0.36.4_emcnas_CD1.iso pkg_6.0.36.4_emcnas.tar

Note: After running the Health Check, be sure to resolve any errors before performing the Software Update.

11



Upgrade Flare 1. Double click the Unisphere Service Manager icon on your desktop.

2. Click Login and log in to your storage system. Enter the storage array admin account name and password and verify that Scope is set to Global.

3. In the menu bar, click Software > System Software and select the Prepare for Installation wizard.

4. In the Prepare for Installation wizard, select the Download software and verify storage environment option, and follow the steps to the Select Software screen. Select the Any supported software option. This sequence of steps allows you to select only the Flare package for downloading, rather than all other associated but optional software items.

5. When the Prepare for Installation wizard is finished, start the Install Software wizard.

6. Follow the instructions to install the desired software packages, for example, Flare 30 release: 04.30.000.5.004. The actual download file is named CX4-Bundle-04.30.000.5.004.pbu.

12



7. When the installation is complete, select Notify your service provider with updated storage system configuration information, and click Finish.

Version 6.0.36.4

The following new features are supported as of Celerra Network Server version 6.0.36.4.

256TB capacity/blade

With the introduction of the Celerra 6.0 software release, we have increased the capacity/blade of the NSG8 and NS960 platforms to 256TB per blade. This enables users to increase the number of disk drives per blade without needing to acquire an additional blade, thereby reducing the user's TCO.

64-bit OS

The Celerra 6.0 release introduces the 64-bit version of the DART operating system. 64-bit technology enables blades with large memory configurations to effectively utilize system resources in order to provide capacity and scalability enhancements across the board.

Control Station auditing system

The system auditing feature has been enhanced with new trigger mechanisms to detect potential system access violations and to start up automatically on single and dual Control Station configurations.

For more information, see the Auditing in the Celerra Control Station Technical Note.

Celerra Upgrade Tool

The Celerra Upgrade Tool (CUT) lets you upgrade your Celerra systems without assistance from EMC or an EMC service provider. CUT communicates over the Internet to EMC Online Support and over a local network to a Celerra system.

CUT is part of the Unisphere Service Manager (USM), and you will now be able to use the same tool to upgrade both Celerra and CLARiiON systems.

CUT uses three wizards to safely guide you, step-by-step, through the process of upgrading the Celerra software.

13



The Celerra Download Wizard selects the latest Celerra software for your system. Software selection is based on your hardware configuration, the software currently installed on the Celerra, and the storage system software running. The wizard can examine your system to obtain this information, or you can provide the information. The Download Celerra Software Wizard gives you one or two choices depending on your configuration, a maintenance release with the latest bug fixes or a new version with new features. The wizard then helps you download the appropriate version of the Celerra software to a central repository on a PC or laptop.

The Pre-Upgrade Health Check Wizard examines the health of the system to determine if the Celerra will support a successful software upgrade.

The Celerra Upgrade Wizard runs the pre-upgrade health check on the Celerra, upgrades the software, and then runs a post-upgrade health check.

Unified Storage system privileges required to run Celerra Upgrade Tool wizards

With the release of NAS 6.0, Unified Storage systems can now participate in a CLARiiON Storage domain as a member system. After becoming a member of the CLARiiON domain, whether the system joined automatically or manually, global user accounts are automatically mapped to the Celerra with Operator (nasadmin) privileges by default. By design, the migrated accounts do not have the necessary privileges to run any of the Unisphere Service Manager's (USM) Celerra Upgrade Tool (CUT) wizards.

USM displays only the Celerra Download and the Celerra Health Check wizards. The Celerra Upgrade wizard is displayed only if the fullnas (nasadmin) or root privileges are assigned to the migrated Users.

In addition, to use all three CUT wizards, the Control Station shell allowed privilege must be enabled for each user. This right allows users to fully use the Celerra Download and the Celerra Health Check wizards. In order to run the Celerra Upgrade wizard, users must also have fullnas (nasadmin) privileges.

To fully use the CUT wizards:

1. Run Unisphere and enter a hostname or IP address.

2. In the login dialog, log in to the Celerra as root.

14



3. From the Scope drop-down list, select Local.

4. Select Settings > User Management > Users. Highlight the user account to which you want to assign privileges and click Properties.

5. In properties, from the Group (Role) Membership field, select fullnas (nasadmin). From the Client Access field select Control Station shell allowed.

Once you have set these privileges for users, they can run USM and log in to a Celerra system and upgrade it with CUT.

Celerra VG2 and VG8

Celerra now supports the new low-end and high-end gateway platforms VG2 and VG8. These two platforms are based on the newest Intel Technology which provides greatly increased performance. The low-end platform, VG2, introduces UltraFlex™ technology for the first time in this level of NAS product and improves the system availability with the option of two Control Stations for high availability failover.

CEPA support for NFS audit applications

This feature provides event notifications and context to audit applications that monitor the NFS file system activity on Celerra, which run in the Windows environment. The audit applications require event notifications, from the Data Mover, to organize the access of information that is stored on the Celerra file system.

To provide this functionality, the Celerra Event Publishing Agent (CEPA) API was modified to allow the applications to obtain the required Data Mover event information without blocking the CIFS or NFS client applications.

To use this feature, the audit applications need to register for post-event notifications through the CEPA API. The CEPA API consists of an IDL file and an XML DTD file. These files contain information that is required by the audit application to interact with the CEPA. The audit application may coexist with the Celerra Event Enabler (CEE) framework on the same client (where Local RPC is used for communication) or on the remote client (where MS-RPC is used for communication). The CEE facilitates the use of selected third-party applications with the Celerra file systems. It provides event values containing the required context as defined by the audit applications for each class.

As more vendor applications are added to each class, the events and associated context are modified to accommodate the applications.

15



The Celerra Event Publishing Agent technical module provides more information.

CLARiiON SATA EFD disk volumes

CLARiiON SATA EFD disk volumes are now supported, in addition to CLARiiON Fibre Channel disk volumes. Celerra support includes the ability for users to provision storage and create / extend file systems on these drives.

The SATA EFD drives are treated identically to FC EFD drives which mean both drive types can be part of the storage pool, one drive type can be replaced with another, and file systems can span FC EFD and SATA EFD drives.

The Managing Celerra Volumes and File Systems with AVM book has more details relating to EFD drives.

Control Station Linux upgrade

As of this release, the Celerra Control Station is upgraded to Red Hat Enterprise Linux 5 (RHEL 5). RHEL 5 leverages new security features, updates, and bug fixes in the kernel and applications provided by RedHat.

Dual port 10GbE SLIC for Data Movers

A new dual-port 10GbE UltraFlex I/O Adapter is being made available for the NS960, NSG2 and NSG8 platforms. This I/O Adapter doubles the effective throughput from the prior generation, which consisted of a single port card. In addition, the UltraFlex form factor allows for easy installation and upgrade to any chassis that supports UltraFlex adapters.

Express Install

Express Install introduces a method to reset the Celerra system back to its default factory configuration.

FAST Cache for the Provisioning wizard

As of this release, Celerra supports EMC FAST Cache as an extension to the dynamic random access memory (DRAM). This extended cache can potentially improve performance of the overall system by increasing write cache size at a comparably low cost.

FAST Cache can be configured anytime by using Unisphere. The provisioning wizard in CSA and UMPW provides you with an option to

16



skip the provisioning of the EFD drives in order to configure them as the FAST Cache drives (maximum 6). You can continue to provision the remaining drives for file / block use.

FLARE R30

Celerra Network Server now supports the latest CLARiiON FLARE release, R30 when installed on CLARiiON CX4 storage arrays. It supports all R30 features, but does not actively support Fully Provisioned Pool LUNs, Auto-Tiering and Compression.

Refer to the E-LabTM Interoperability Navigator at

http://Support.EMC.com for more specific FLARE version numbers.

IPv6

IPv6 is fully supported and configurable by means of the CLI, and partially supported (read-only and some limited configuration) by means of Unisphere.

CLI commands

All CLI commands now support IPv4 and IPv6 addresses. There are some new commands provided specifically to handle IPv6. Some existing commands have updated syntax to support IPv6 addresses. Supported areas include: NFSv2/v3, FTP, TFTP, NIS, DNS, partial LDAP, Kerberos, CIFS, iSCSI, Network, NTP, NFSv4, Secure NFS, UserMapper, ReplicationV2, XLT, CDMS, FileMover, Control Station, SNMPv1 traps, SNMPv3, MirrorView, SRDF, CAVA/VirusChecker, ConnectEMC and Celerra XML API V2. The Configuring andManaging Celerra Networking technical module provides detailed information about IPv6 networking.

IPv6 read-only support

Unisphere provides IPv6 read-only support with a few exceptions. If IPv6 is configured by means of the CLI, that configuration is visible on the Unisphere. The information about auto-generated link-local interfaces and routes is not visible by means of Unisphere.

The IPv6 addresses in drop down lists and checkboxes display IPv6 configuration information and are selectable. IPv6 addresses in hyperlinks navigate to the appropriate destination. Some screens allow the selection of IPv6 address for configuration in CIFS and Replication. The netmask field displays either an IPv4 subnet mask for IPv4 addresses or a non-editable IPv6 prefix for IPv6 addresses.


17



The broadcast address is blank for IPv6. The IPv4 network configurations are fully supported. When there are combined configurations of IPv4 and IPv6, changes via Unisphere are prohibited.

Unisphere wizards, including the CIFS, Control Station, iSCSI, Network, Replication, and Set up Celerra wizards provide read-only support for IPv6. Lists containing IPv6 address are selectable, but not editable.

CBM applications Client-side CBM applications may not work with an IPv6 configured iSCSI LUN.

Celerra Monitor Celerra Monitor provides read-only support for IPv6. When IPv4 is configured, the screen displays the IP address, netmask, and broadcast fields. When IPv6 is configured, the screen displays IP address and the subnet prefix length fields.

CSA The Celerra Startup Assistant (CSA) does not provide IPv6 support at this time. A new CLI command (nas_cs) provides IPv6 support.

IPv4-only or IPv4/IPv6 environment? Should you configure IPv4-only environment or a dual stack IPv4/IPv6 environment? When deciding this, you need to answer the following question:

In my environment, do any nodes or peers served by the Celerra use the IPv6 protocol?

If you answer yes, the Celerra and associated infrastructure, such as routing, Domain Name System (DNS), and Network Information Service (NIS), should be configured for a dual stack environment.

Link-local information The Celerra XML API V2 supports IPv6, however the auto-generated link-local information is not visible for the XML API V2.

The application protocols are not supported via link-local interfaces. The link-local related information is supported only for IPv6 core protocols such as Duplicate-Address-Detection, Router Discovery, Neighbor Discovery, and ICMP6 and is exposed only via the server_ping6 command at this time. Only the CLI will have the view of all the IPv6 interfaces.

MMC Microsoft Management Console (MMC) Snap-in provides full IPv6

18



support. That is, MMC accepts both IPv4 and IPv6 addresses that you input. Machines hosting MMC must be running Vista or Windows Server 2008 or later to support IPv6.

Kerberos authentication support for LDAP-based directory services

In this release Celerra adds support for Kerberos authentication to the LDAP-based directory services provided by Active Directory with Microsoft Windows Services for UNIX (SFU) or Identity Management for UNIX (IdMU).

When configuring LDAP, the Data Mover’s directory client can now use Kerberos, simple (password), or SSL authentication. Anonymous authentication is not allowed.

Kerberos authentication means the Data Mover, configured as a CIFS server, uses a KDC to confirm the Data Mover’s identity when accessing the Active Directory.

The Configuring Celerra Naming Services technical module describes how to configure a Data Mover which is configured to use Active Directory with SFU or IdMU to use Kerberos authentication.

MAC OS 10.6

In this release, EMC qualified MAC 0S 10.5 and 10.6 in a CIFS protocol environment. File access protocol using native CIFS protocol stack in MAC-OS client has been verified in term of compatibility with Celerra CIFS implementation. The 'AppleSingle' feature which is using data stream to store resource information instead of a second file associated to the data file has also been validated (with MAC-OS 10.6, this is the default behavior on the client mount option).

MPFS Operating Systems

MPFS now supports Windows 7 clients (x86 and x86_64), Windows 2008 R2 (x86_64 only), CentOS 5.4 (x86 and x86_64), SLES 10 SP3 (x86, x86_64) and SLES 11 (x86 and x86_64 only). CentOS is supported over MPFS-iSCSI only. All other operating systems are supported over Fibre Channel and iSCSI. For Windows 7 and Windows 008 R2, MPFS supports SMB1, SMB2 and SMB2.1 versions of the CIFS protocol.

19



NDMP

Increased NDMP backup streams

The increased NDMP backup concurrent streams feature enables the Celerra Network Server to increase the maximum number of simultaneous concurrent backup sessions from four to eight. It verifies that the system memory is at least 8 GB, and then sets the value to eight. If the system memory is less than 8 GB, the value defaults to four simultaneous sessions. The Celerra Network Server Parameters Guide provides more detailed information about this feature.

To support eight concurrent backup or restore streams:

The backup client (DMA) must be compatible and configured to run eight concurrent streams.

The Data Mover must have at least 8 GB memory.

Increasing the number of concurrent sessions to eight may result in an increase in the utilization of CPU and system memory. The Configuring NDMP Backups on Celerra technical module provides more information.

NDMPcopy

The NDMPcopy feature is an open source program that allows users to copy file system data within the Data Mover or between different Data Movers that are on the same or different Celerra.

The NDMPcopy enables users to directly transfer data from the source to the destination file system, without traversing the NDMPcopy host machine or network (if the source and destination file systems are on the same Data Mover). It uses NDMP (Network Data Management Protocol) to transfer the file system data.

With the NDMPcopy program, the NDMP host and network (if the source and destination file systems are on the same Data Mover) are not overloaded, because the data transfers take place directly from the source to the destination. The NDMPcopy program contacts the NDMP server processes on the source and destination file systems, and authenticates the connection. It then initiates data transfers between the source and destination file systems. The NDMP host receives status messages only.

The NDMPcopy also provides multiprotocol (NFS/CIFS) support. EMC recommends using the NDMPcopy program for data transfers only. The NDMPcopy may affect the scheduling of backup activities in the same source, as it modifies the BackupDates file on the Data Mover. It uses the text mode for authentication and so the password is sent across the

20



network in a non-encrypted format.

The Configuring NDMP Backups on Celerra technical module provides more information.

NDMP port range

The NDMP port range feature allows users to configure the port range to use a limited range of TCP ports that correspond to the ports for which the firewall allows inbound connections. EMC recommends setting the NDMP port range using dynamic or private ports (49152-65535) of IANA. The required minimum value of the port width is 32; however, EMC recommends setting the width to at least 100. Setting the port range does not imply that these ports are reserved for NDMP connection.

When the port range is set, NDMP selects a random port number from the specified port range as the data connection listening port. If the port range is not set, the default port range is used.

NDMP is incompatible with NAT (Network Address Translation) firewalls. It uses the existing server_param management interface and is independent of the DMA. Configuring the NDMP port range allows the Data Mover to use a limited range of TCP ports that correspond to the ports for which the firewall allows inbound connections. The Configuring NDMP Backups on Celerra technical module and the Celerra Network Server Parameters Guide provide more detailed information about NDMP port range.

pNFS

With this release, the Celerra Network Server provides a conforming implementation of NFSv4.1 in order to support pNFS. The pNFS is a part of NFS version 4.1 that allows clients to access storage devices directly and in parallel. In order to support pNFS, the Celerra Network Server fully supports all required NFSv4.1 functionality. The pNFS architecture enhances the scalability and performance of the existing NFSv4 servers. This is achieved by the separation of data and metadata, and moving the metadata server out of the data path.

Quotas

Choose and set a quota policy

As of this release, administrators can change the quota limit check policy without turning off (disabling) all of the quotas on the file system. The new policy change will be applied after the Data Mover is rebooted.

21



The technical module Using Quotas on Celerra provides additional information about this change.

RSA Common Security Toolkit module

In this release the RSA Common Security Toolkit (CST) module replaces the CAMS Common Sign-On (CSO) module which was previously used to handle remote authentication of Celerra administrative users against an external directory server.

CST changes the operation of the directory server settings test function. The Celerra Security Configuration Guide and Unisphere Settings > User Management online help provide more information about this feature.

Samba 4

Celerra supports Samba 4, except for the Samba 4 domain controller. Samba is an open source / free software suite that provides file and print services to all manner of SMB/CIFS clients, including the numerous versions of Microsoft Windows operating systems.

SAN Copy

SAN Copy™ replication is an array based replication technology. The performance and scalability of array based solutions is better for large-scale migration than host based applications. SAN Copy is integrated into the Unisphere console and simplifies the configuration and management of LUN migrations through a user friendly GUI interface.

The SAN Copy CLI Reference Guide and the on-line help contain additional information about SAN Copy.

SMI-S v1.4

The new SMI-S feature enables generic SMI-S compliant clients to manage a Celerra using CIM (Common Information Model) operations defined for NAS.

This feature delivers two components of WBEM (Web Based Enterprise Management) infrastructure; namely a CIM Server and a Provider. Both components are embedded on the Celerra Control Station. SMI-S contains an API that storage management applications can use to provide standardized view of a Celerra. Providing a compliant API in SMI-S terms means supporting profiles and sub-profiles of the NAS domain. For Celerra, this means supporting the NAS Head Profile at a minimum, along with its mandatory subprofiles.

22



SNMPv3

The Simple Network Management Protocol (SNMP) collects and processes valuable network information. It gathers data by polling the devices on the network from a management station at fixed or random intervals.

In this release, a new SNMP agent, the SNMP daemon (SNMPD), which runs on the Data Mover, supports SNMPv1, SNMPv2c, and SNMPv3. SNMPv3 supports IPv4, IPv6, and enhanced security over SNMPv1 and SNMPv2c.

SNMPv3 provides important security features such as:

Authentication to verify that the message is from a valid source

Encryption of packets to prevent snooping by an unauthorized source

Message integrity to ensure that a packet is not tampered with in transit

Tolerance for 8 GB FC in Symmetrix

When a Celerra gateway with 4G Fibre Channel (FC) is attached through a switch to a Symmetrix platform incorporating an 8 GB FC card, there is no loss of or change in functionality. The communication rate automatically drops to 4 GB (Symmetrix platform data rate negotiation) and operation continues as expected.

Tools

Windows event log auto archive

With Windows operating system, applications can use the event logging mechanism to log their own events. VNX currently supports three such event logs that are security, system, and applications.

The physical format of these logs use a Microsoft format called 'evt' that has a limitation of 4 GB in size because there are some fields stored on 32 bits integers. Windows 2008 has introduced a new format 'evtx' that does not have this limitation.

The event log auto archive feature allows you to automatically archive an event log on a particular trigger and to continue the logging on a new event log without losing any events. The archive is triggered on a time and/or on an event log size basis defined by parameters in the Windows registry. You can specify a retention policy to keep the event log archives before they can be recycled based on the duration and/or the total archive disk size.

23



All parameters are stored in the Windows registry of each VDM. Therefore, each VDM will have its own configuration. The parameters can be viewed and edited with tools like regedit. The Configuring and Managing CIFS on VNX technical module provides more information about this feature.

Unisphere

Version 6.0.x introduces Unisphere. Unisphere is a unified SAN/NAS application providing uniform and intuitive management and monitoring services. It replaces Celerra Manager, Navisphere Manager, and Navisphere Express.

Highlights of the new Unisphere application include:

Single sign-on for a user identity common to multiple Celerra and CLARiiON systems. This functionality is facilitated by creating a global domain user who can also manage Celerra systems. This common user only needs to log in to one system in the domain in order to access and manage all systems.

Management of up to 50 Celerra and CLARiiON systems within a single Unisphere instance.

System summary views in a new dashboard that provides alerts and status summaries for the managed system list populated by the user.

Participation in storage domains by both CLARiiON and Celerra systems.

Cross-linking capability for LUN, disk group, storage pool, and storage system properties from Celerra to the relevant CLARiiON.

Note: All existing Celerra features, with the exception of message-of-the-day and system monitoring after session timeout, are supported within Unisphere.

The following table lists where each Celerra Manager view can be found in Unisphere. In some cases, Celerra Manager views map only to tasks found in Unisphere’s left-pane task list.

24



Table 1 Celerra Manager to Unisphere Feature Mapping

Celerra Manager Unisphere Celerra Home > Status All Systems > Dashboard > Alerts Celerra Home > Licenses System (System tasks) > Manage Licenses Celerra Home > Statistics Monitoring (Statistics tasks) > All Performance Celerra Home > System Info System > Details Celerra Home > Control Station Properties

System > Details System (System tasks) > Manage Control Stations

Data Movers System > Data Movers Data Movers > Virtual Data Movers Sharing > CIFS

Data Movers > Properties > Notifications Monitoring > Notifications Storage > Systems System > Details Storage > Systems > Properties System > Details Storage > Systems > Properties > Disk Groups

Storage > RAID Groups

Storage > Systems > Properties > Spare Disk Devices

Storage > Hot Spares

Storage > Volumes Storage > Volumes

Storage > Pools Storage > Storage Pools Storage > Pools > Usage Statistics

Storage (Storage Pools tasks) > View Usage Statistics Monitoring (Statistics tasks) > Storage Pool Usage

File Systems Storage > File Systems File Systems > Properties Storage > File Systems > Properties File Systems > Properties > Checkpoint Schedules

Storage > File Systems > Properties

File Systems > Properties > Mounts Storage > File Systems > Mounts File Systems > Properties > Checkpoint Storage

Storage > File Systems > Properties

File Systems > Properties > Quota Settings

Storage > File Systems (File Systems tasks) > Manage Quota Settings

File Systems > Properties > Notifications Removed File Systems > Mounts Storage > File Systems File Systems > I/O Statistics

Storage (File Systems tasks) > View I/O Statistics Monitoring (Statistics tasks) > File System I/O

25



File Systems > Usage Statistics

Storage (File Systems tasks) > View Usage Statistics Monitoring (Statistics tasks) > File System Usage1

File Systems > Deduplication Settings File Systems > Properties

File System Quotas Storage > File Systems File System Quotas > Settings Storage > File Systems (File Systems tasks) >

Manage Quota Settings

SnapSure Checkpoints Replicas > Checkpoints Data Migration Storage > Data Migration CIFS Sharing > CIFS CIFS > Statistics Sharing (CIFS tasks) > View Statistics

Monitoring (Statistics tasks) > CIFS

NFS Exports Sharing > NFS NFS Exports > Statistics Sharing (NFS tasks) > View Statistics

Monitoring (Statistics tasks) > NFS

iSCSI Sharing > iSCSI iSCSI > Configuration Sharing (iSCSI tasks) > Manage Settings Replicator Replicas > Replications Network System > Network Network > NIS Settings System (Network tasks) > Manage NIS

Settings Network > Ping System (Network tasks) > Run Ping Test

Network > Statistics

System (Network tasks) > View Statistics Monitoring (Statistics tasks) > Network Device

VTLUs Storage > Virtual Tape Events Monitoring > Event Logs Notifications Monitoring > Notifications

Inventory System > Hardware Task Status Monitoring > Background Tasks CLI Commands All task lists Security > Administrators Settings > User Management Security > Celerra Manager Settings (task list) > Manage Idle Timeout

Security > Network Services System > Network Security > Public Key Certificates Settings > Public Key Certificates Tools > Celerra Monitor Monitoring (task list) > Launch Celerra Monitor Tools > Launch Navisphere n/a Wizards – Network System (Network tasks)

26



Wizards – CIFS Sharing (CIFS tasks) Wizards - File System Storage (File Systems tasks)

Wizards – Replication Replicas (Replications tasks) Wizards – iSCSI Sharing (iSCSI tasks) Wizards – Celerra System (System tasks) Support System (Service tasks)

Upgrade media

Dual Compact Disk

As of this release, the upgrade media is delivered on dual CDs. An out-of-family upgrade, for example, V5.6 to V6.0 requires both CDs. The software upgrade prompts the user to swap CDs when necessary.

DVD

The upgrade media is available on one DVD for users of Control Stations that have optical drives capable of reading DVD discs.

Users and roles

Version 6.0.x introduces several changes to the Celerra user and role-based access feature.

CLI role-based user access A user account is always associated with a primary group and each group is assigned a role. A role defines the privileges (that is, the operations) the user can perform on a particular Celerra object.

In version 6.0.x, Celerra extends support for assigning role-based privileges to a user accessing Celerra through the CLI. The ability to select a predefined role or define a custom role that gives a user the necessary privileges to execute a command is not available for all CLI commands. Certain commands can only be performed by the default user accounts root and nasadmin or, in some cases, by a user account associated with the root and nasadmin roles.

You create and manage role-based user access in Unisphere Settings > User Management > Roles. You must be root or a user who has root or security operator privileges to create a user account and to associate it with a group and role.

The Celerra Security Configuration Guide, Celerra Network Server Command Reference, and Unisphere Settings> User Management online

27

Known problems and limitations


help provide more information about the specific command actions for which privileges can be controlled.

Unisphere storage domain users and roles

To simplify access and privileges for users managing both CLARiiON and Celerra systems in Unisphere, user accounts created to manage CLARiiON storage are mapped to Celerra user accounts, groups, and roles so that CLARiiON storage users can be given privileges to access and control Celerra objects. Groups and roles with the prefix sd_listed in Unisphere Settings> User Management views are mapped from CLARiiON storage domain groups and roles.


CDMS

Connecting to a Solaris secondary server

Users are unable to create a CDMS or a FileMover connection to a Solaris secondary server using TCP with IPv6. IPv6 with NFSv3 will not be able to create this connection.

If UDP is chosen instead of TCP, this type of connection will work with a Data Mover NFS server. This problem has only been seen with Solaris NFS servers.

Workaround Specify UDP as the transport protocol instead of using the default TCP. This option can be set with both the server_cdms and fs_dhsm commands.

Celerra Event Enabler

The following known problems and limitations pertain to Celerra Event Enabler:

CAVA/CEPA servers

Each Data Mover should have a CAVA pool consisting of a minimum of two CAVA servers specified in the Data Mover’s “viruschecker.conf” file, or a CEPA pool consisting of a minimum of two CEPA servers specified in the Data Mover’s “cepp.conf” file.

28



CIFS and NFS clients

The CAVA feature of the CEE product is for CIFS and NFS clients. If FTP protocols are used to modify files, the files will not be scanned for viruses. No UNIX antivirus application support is provided.

File-Level Retention

It is strongly recommended that the antivirus (AV) administrator updates the virus definition files on all resident AV engines in the CAVA pools, and periodically run a full file system scan of the file system to detect infected FLR files. The Using Celerra AntiVirus Agent technical module provides additional information about virus definition updates.

Microsoft Access

Do not scan Microsoft Access database files in real time. Files inside a database are not scanned.

Windows 64-bit operating systems

In order to run CEE on Windows 64-bit operating systems, the Celerra-to-CEE communications must be over Microsoft RPC (MS-RPC). The version of CEE that runs on Windows 64-bit operating systems is supported with Celerra version 5.6.45 or later and CEE version 4.5.0.4 or later.

Celerra iSCSI AIX

A known problem with SMIT (System Management Interface Tool) in AIX versions 5.2, 5.3, and 6.1 prevents it from creating and handling compliant IQN names for iSCSI nodes.

As a workaround, manually edit the file /etc/iscsi/targets to enter or modify IQN names. Check the IBM support website for additional information.

Configurations

Celerra iSCSI configurations in the backend replication environment do not support TimeFinder®/FS, NearCopy, and FarCopy.

Host connectivity issues

Check the following guides for issues related to the Microsoft initiator, QLogic HBAs, and compatibility with the Windows operating system:

Host Connectivity Guide for Windows

Host Connectivity Guide for Linux

29



These guides are available from EMC Online Support at http://Support.EMC.com.

Microsoft iSCSI Software Initiator

EMC Celerra iSCSI supports the most current version of the Microsoft iSCSI Software Initiator, which is available as a free download from Microsoft’s website (search for iSCSI initiator). This section contains known issues with the Microsoft iSCSI Software Initiator.

Initiator uses only the first network portal of a target

An iSCSI target on the Celerra Network Server may have multiple network portals configured for it; however, the Microsoft iSCSI initiator GUI uses only the first network portal of a target.

Note: Applies only to Microsoft iSCSI Software Initiator versions 2.0 and earlier.

Workaround Use the Microsoft initiator CLI to access other portals.

Control Station

Authentication failure

Following a dual Control Station upgrade to Celerra Network Server version 6.0, users can log in successful using LDAP on the primary. When the Control Station fails over, the LDAP login no longer works. The following error is displayed: Session timeout. Relogin and try this operation again. This issue only affects LDAP users; local root/nasadmin accounts and global accounts are still available.

Workaround

Run /nas/site/cst_setup -reset to resync the CST settings. Once those settings are synced, you will be able to failover/failback and successfully be able to login each time using LDAP.

Configuring LDAP domain settings

When the Control Station is freshly installed and the Custom Active Directory page is used to configure the initial LDAP domain settings, the distinguish name is not set properly. Testing the Domain Settings generates an error similar to the following: Primary server 12.3.45.67 bind error. This issue only occurs during the initial setting. Once the


30



initial setting is done, the issue never happens again.

Workaround Use the 'Default Active Directory' option to configure the initial LDAP domain settings and then use the other options to change the settings.

Copying large files to remote over NFS

When a user copies very large files from the Control Station to a remote file system over NFS on a slow network, other users of that Control Station may experience slow performance when running CLI commands. For dual Control Stations, this issue can cause the Control Stations to fail over. Over time, the Control Station will correct itself, and performance will return to normal.

Domain attributes

After a Control Station failover on a Celerra that is a member of a storage domain, the administrator will not be able to use or see domain changes made while the Celerra was in the CS failover state.

Workaround Continue using the Celerra with the domain attributes it currently has, or remove the Celerra from the domain and re-add it.

Linux automount

The /nasmcd/quota and /nasmcd/rootfs directories are used to nfs mount the root FS of the Data Movers using the Linux automount package. Occasionally, automount's attempts to perform this nfs mount fail, resulting in the requested Data Mover file system not being available. This problem happens more frequently when there are multiple automount requests for the same share over a short period of time. When the mount does fail, retrying the attempt after a short delay will usually be successful.

One area that may be impacted by a mount failure is quota management, as it relies on access to files under /nasmcd/quota to retrieve quota information. The quota functionality and other operational aspects of the Data Movers are not affected.

Security auditing

When security auditing is stopped for any reason, for example, due to a lack of disk space, no alert is raised to the administrator.

31



Deduplication

Deduplicated data space not reclaimed

A race condition occurs between removal of deduplicated files and file system unmount operations. Under certain conditions, deduplicated data space may not be reclaimed from the file system when files are removed. Messages similar to the following are displayed in the server_log when this issue occurs:

2010-05-09 09:40:57: 13163036672: DEDUPE: 3: Cannot find file system with id 15, cannot cleanup object <dart://rde..."

Free Space

During the deduplication process, the file system must have enough free space available that is equivalent to the size of the original file to be deduplicated plus the size of the compressed version of the file to be stored. An additional 1 percent of the file system must be free, or if auto-extension is enabled, an additional 1 percent below the auto-extension threshold must be free.

Restoring space-reduced files

Restoring space-reduced files from a PAX-based NDMP backup into a file system causes the deduplication state to be set to suspended if the deduplication state is not currently enabled.

I/O errors reported or deduplicated data space not reclaimed

A race condition occurs during domain re-initialization. Under certain conditions, I/O errors can be reported or deduplicated data space may not be reclaimed from the file system when files are removed. Messages similar to the following are displayed in the server_log when this issue occurs:

removeFromSecondaryStorage could not determine ssp

File-Level Retention

Prior to Celerra version 5.6.43.9, FLR protected empty locked files from being deleted. As of version 5.6.43.9, both FLR-E and FLR-C allow the deletion of empty locked files and empty append-only files. Any locked or append-only file that contains data will be protected.

This behavior will be changed in a future release such that all locked and append-only files will be protected regardless of whether they contain

32



data.

File system

Mounting file systems

When mounting a file system as read-only on two (or more) Data Movers, the following error message appears on the copy file system pop up: "There are no Data Movers interconnect available on server_3, server_2".

The error message is misleading and can be ignored. EMC does not support creating a copy of a file system that is to be mounted to more than one Data Mover.

nas_fsck command

A race condition occurs when using nas_fsck command on file systems smaller than 2G. This causes the file system to be unavailable to the user.

Timestamp mismatch

After a restore operation, a timestamp mismatch occurs on the softlinks with file names longer than 83 characters. The timestamp shows current time (i.e. time of restoration) instead of the original timestamp on the link.

Installation and upgrades

Celerra now joined to domain during upgrade

The Celerra is now being joined to the CLARiiON domain automatically during an upgrade. You can use Unisphere to elevate the privileges of any user that has been transferred from the Domain to the Celerra. This does not occur for gateway machines, however. For gateways, users will need to join the Celerra to the domain they want it to be a part of.

Celerra Startup Assistant

After a fresh factory Express Install of an NS40FC system, the CSA fails on the Pre-Configuration apply changes page with the following error: CLARiiON's Domain Join Setup | Command Failed.

If this occurs, ignore the error; this step should not have been tried, and failure poses no operational problems.

33



CallHome messages

The Celerra uses connectEMC 2.6 software to deliver callhome messages to EMC. The outgoing mail functionality of this version issues duplicate HELO/EHLO commands. This is in accordance with RFC 2821, however, some mail servers do not support this (especially older mail servers). If this occurs in your environment, consider updating your mail server.

Control LUN requirements enforced

There is strict enforcement of the control LUN size and HLU numbers for fresh installations. An exact match between the LUN capacity and the correct HLU value is required. Mismatches will generate a "Control LUN Check Failed" error message during installation. You must correct the control LUNs before installation can complete successfully.

Determining NAS version

While using CUT to perform an upgrade from 5.6 to 6.0, the CUT tool presented a pop up saying it could not determine the nas_version, after or during the Linux upgrade portion of the upgrade. Retrying the task succeeds.

Error message

While upgrading from version 5.6 to 6.0, the following error message is displayed: Error - cannot open /usr/sbin/ext2online /dev/hda1.

This error can be ignored, as it does not stop the upgrade and the task completes successfully.

Gateway and FC-enabled systems

During upgrades of gateway and FC-enabled systems, a message stating that the Celerra was joined to the back-end domain is presented. That message is incorrect, and can be ignored. The Celerra can be joined to the domain using CLI commands or via UEM on the domain master system.

Incorrect certificate

Users may experience an invalid certificate after fresh installing their system.

Workaround Set the correct time on the Control Station and regenerate the certificate, or wait for the certificate to become valid and login again.

34



nasadmin credentials

If you pause and resume the upgrade, the "Clariion domain nasadmin Credentials" screen is not displayed.

Workaround Use the nas storage command to change the password of nasadmin.

nasadmin password

When logging in globally and attempting to change the nasadmin password, an authentication error is generated and users are not allowed to change the password. The workaround is to use a nas storage command to change the password of nasadmin.

Password confirmation

The "Clariion domain nasadmin Credentials" screen in the Celerra Upgrade Tool (CUT) currently does not have a confirm password field. Users can mistype the password without knowing it, and will be unable to log in.

PUHC "Check for running tasks" fails

While upgrading from version 5.6 to 6.0 using USM, the PUHC "Check for running tasks" fails and offers a set of actions to take. These actions do not work. The first step says to type the following; "<br> * nas_license -list". Neither of the 2 steps offer strings work if typed in as specified and produce errors.

Supported hardware

When attempting to run USM on unsupported hardware (CNS|CFS|NS350|NS500|NS600|NS700), the Celerra Upgrade Tool will correctly exit but exit code is not passed back to the USM. Therefore, the USM tries to continue the upgrade process but will fail on the next step that interacts with the Celerra system.

The error returned at that point does not give a clear indication of what the actual failure was. If you see unclear error messages while using USM, make sure that you are using a supported hardware platform.

Upgrade paused

During an upgrade with the CLI, and when it was paused during the Control Station reboot, the Upgrade wizard and on the welcome screen tells user that the upgrade is paused and allows user to resume the

35



upgrade. Do not use the wizard to resume an upgrade that is started in the CLI.

Multibyte support

Using multibyte characters in fields in which they are not supported. If you use multibyte characters in data element fields that do not support multibyte characters (for example, checkpoint schedule descriptions), these fields will not display correctly. If you want these fields to be readable, you will have to recreate or rename the existing data element text using the ASCII character set.

Refer to PRIMUS emc224755 for more detailed information about this limitation.

Using multibyte characters with ECC

Celerra version 5.6.40.3 and later may cause discovery and name display issues for ECC 6.x if certain Celerra object names use non-ASCII characters. The relevant names are:

File system name

Mount point name

CIFS share name NFS export name

Quota user name

Discovery issue If non-ASCII characters are used and if the name is more than 63 characters, discovery may not work.

Garbled name display issue If these names use non-ASCII characters, the names do not display properly in ECC. However any ASCII characters used in the name do display properly. Furthermore, the name itself and the underlying data on the Celerra system are not affected.

Suggested Workarounds Use only ASCII characters in the specified object names when using

ECC for Celerra discovery. If multibyte characters must be used, keep all applicable object names under 63 characters.

Since ASCII characters display properly, append or prepend ASCII characters to object names so you can use these characters to find the correct name.

36



Refer to Primus emc229599 for more details about this issue.

NDMP

Incremental NDMP backups and NDMPcopy operations

Incremental NDMP backups and NDMPcopy operations use the same last backup timestamp tracking for each data set on a Data Mover. A data set in this context means the point where the backup or copy operation starts (e.g., the root of a file system or a specific sub-directory within a file system). This means that if incremental NDMP backup and NDMPcopy operations of the same data set (i.e., starting from the same point within a file system) are interleaved, then the following will occur:

The incremental backup will contain only changes in that data set since the NDMPcopy operation instead of since the last backup operation.

The incremental NDMPcopy of that data set will copy only those files changed since the NDMP backup instead of since the last NDMPcopy operation.

To avoid this, do not interleave NDMP backup operations with incremental NDMPcopy operations of the same data set; if you do, perform a full NDMP backup of that data set after any NDMPcopy activity.

Vendor support

Not all NDMP vendor versions support this version of Celerra Network Server. For the latest NDMP Vendors/Versions/Patches and restrictions supported in this version of Celerra Network Server, refer to the Backup section of the E-Lab Interoperability Navigator.

Network

Do not disable IP reflects when using IPv6. System may panic when using IPv6 if IP reflect is disabled.

NFS

Cached files

If a client mounts exported parent and child paths of the same file system, it is possible to access the same file from two different mount points. If client side caching is enabled, the client may end up maintaining two different versions of a cached file.

37



Celerra Data Migration System

Access to online migration file systems while using CDMS is not supported by NFSv4 clients until migration completes.

When NFSv4 is used, it triggers directory migration of subdirectories and requires global lock for the entire Data Mover. Therefore, the Data Mover will be inaccessible during the migration.

NFS export failure

If the export of a corrupted file system fails, the export entry of the file system which was corrupted & cleaned using FSCK is not seen in the export list. However, if user manually exports it again, the entry can properly be seen in the NFS export.

Quotas

Performance statistics

Termination summary is not sorted for Set statistics if a statpath name is duplicated in the server_stats command. The following example will not have a sorted termination summary: server_stats server_x -monitor store.diskVolume,store.diskVolume -sort writes.

This issue does not impact the sorting of the sampled data displayed periodically during the server_stats session, only the termination summary at the end of the session is impacted.

Workaround Not using duplicate statpath names when -sort is specified will eliminate this issue. For example: server_stats server_x –monitor store.diskVolume -sort writes.

Replication

nas_copy command

When using Replication V2 and the nas_copy command, there is a window where if the X-Blade reboots, results may not be properly reflected. The transfer will complete but the task is not cleaned up and the appropriate status is not reported back to the user. Once completed, the task can be manually aborted.

38



Security

Changing LDAP settings

When running the server_netstat command, or updating the LDAP settings, the following error was displayed:

The lockbox stable value threshold was not met because the system fingerprint has changed. Reset the SSV values by running "/nas/sbin/cst_setup -reset" from the CLI to resolve the issue.

SnapSure

Parameter file

Do not insert a new-line character at the end of a nas_param file (/nas/site/ or /nas/sys/). The empty line causes fs_ckpt to return an encodeable.extract error.

Writable checkpoints operations

Writable checkpoints do not support the following operations and features:

Checkpoint refresh Checkpoint schedule

Celerra replication

CDMS

File-level retention

File system extension

iSCSI MirrorView/SRDF

TimeFinder/RDF

VTLU

Data mover panic (425057)

The data mover panicked. The following error was displayed:

DART panic/fault message: >>PANIC: I/O not progressing LastVol touched Sn891 Kind 5 (ptr=0x65d74b08)

Workaround Contact your service provider.

39



SRDF/S

nas_rdf –activate

The nas_rdf –activate –reverse command may fail to swap device-pair personalities during the activate. This occurs when the source Celerra has a TimeFinder/FS snapshot/s created on R1BCVs which use TimeFinder Emulation mode (Symm 8 supports Emulation mode only). The error message below is specific to the personality swap and having remotely protected storage, and does not impact the successful activation of the destination Celerra system.

Error message in nas_rdf –activate –reverse:

An RDF 'Swap Personality' operation execution is in progress for device group '1R2_500_3'. Please wait...

Cannot use the device for this function because it is a Copy Session target

ERROR 13431799810: Swap operation of device group 1R2_500_3 failed due to the improper backend state or configuration.

Run 'nas_message -info 13431799810' to get full description and recommended action.

Workaround Canceling the relationship between the R1BCV mirrored volumes and the R1STD volumes is needed to complete the swap operation.

After the SRDF restore completes, the SYMAPI will automatically request a full establish during the subsequent fs_timefinder –M refresh command to the snapshots built on the canceled R1BCV volumes.

Table 2 nas_rdf –activate –reverse procedure

Step Action 1 To check symm id of source Symmetrix, run

/nas/symcli/bin/symcfg list on the destination SRDF system.

2 To check the mirrored volumes, run nas/symcli/bin/symmir –sid <source symm id> list on the destination SRDF system. You can also double check if Emulation mode is used from the output.

3 To find R2 device group, run /nas/symcli/bin/symdglist on the destination SRDF system.

40



4 To find R1BCV volumes, run /nas/symcli/bin/symrdf –g <r2 group name> query –bcv on the destination SRDF system. If the same device ID is seen in the output that is the R1BCV device you need to cancel.

5 To prepare for the cancel operation, create a file which has all R1BCV volume pairs to cancel on destination SRDF system.

Example of the file:

# cat file

# <standard device id><bcv device id>

0068 00B8

0069 00B9

6 To cancel the R1BCV volumes, run /nas/symcli/bin/symmir –sid <source sid> -f <file name> cancel on the destination SRDF system.

7 To confirm the canceled volume, run /nas/symcli/bin/symmir –sid <source symm id> list on the destination SRDF system. The canceled volumes should not be listed in the output.

8 To compete the activate reverse operation, run /nas/sbin/nas_rdf –activate –reverse on the destination SRDF system.

nas_rdf –restore

The nas_rdf –restore command may fail on SRDF system which uses TimeFinder Emulation mode (Symm 8 supports Emulation mode only) with R2BCV mirrored volumes. Error message in nas_rdf –restore:

An RDF 'Failover' operation execution is n progress for device group '1R2_500_3'. Please wait...

Cannot use the device for this function because it is a Copy

Session target

CRITICAL FAULT: SRDF restore failed for 1R2_500_3

Workaround Canceling the relationship between the R2BCV mirrored volumes and the R2STD volumes is needed to complete the restore operation.

41



After SRDF restore completes, full establish is required when we do mirror on again to the canceled R2BCV volumes.

Table 3 nas_rdf –activate –restore procedure

Step Action

1 To check the mirrored volumes, run /nas/symcli/bin/symmir –sid list on the destination SRDF system. You can also double check if Emulation mode is used from the output.

2 To find R2 device group, run /nas/symcli/bin/symdg list on the destination SRDF system.

3 To find R2BCV volumes, run /nas/symcli/bin/symrdf –g <r2 group name> query –bcv on the destination SRDF system. If the same device ID is seen in the output, that is the R2BCV device you need to cancel.

4 To cancel the R2BCV volumes, run /nas/symcli/bin/symmir –g <r2 group name> cancel -force on the destination SRDF system.

5 To confirm the canceled volume, run /nas/symcli/bin/symmir –sid list on the destination SRDF system. The canceled volumes should not be listed in the output.

6 To complete restore operation, run /nas/sbin/nas_rdf –

restore on the destination SRDF system.

Nas_rdf –init command fails

Problem: During the SRDF discover remote storage devices process, the nas_rdf -init command fails and returns the following error: "Unable to setup remote devices".

Workaround:

Delete the device group for R2 devices using /nas/symcli/bin/symdg delete <DgName> -force.

For example: /nas/symcli/bin/symdg delete 1R2_500_13 –force. Then run nas_rdf –init, to re-create the R2 device group.

42



SSD drives not supported

SRDF/S does not support solid state disk (SSD) drives. If you attempt to configure an SSD drive with SRDF/S, Celerra health check operations reject the device as an unsupported storage.

Unisphere

The following notes apply to Unisphere.

Adding a system to a Celerra-CLARiiON domain

When users attempt to add a system to a Celerra-CLARiiON domain and uses a LOCAL scope to authenticate to the remote system, it appears to work. However, the system does not actually get added to the domain. The correct procedure is to authenticate to the system to be added with a GLOBAL scope user that is part of the root group.

Adding a system to a Celerra-Symmetrix domain

When attempting to log into a Celerra that has a Symmetrix backend and is part of a storage domain, you may see the following message on the console: "Target system already managed." If this occurs, log out, wait 15 minutes, and log in again or close the domain master page and restart your browser session.

Adding a system to a second domain

Adding a Celerra to one domain followed by adding it to a second domain from a non-master CLARiiON may result in the Celerra appearing to be in both domains at the same time, depending on which system scope you are in.

Workaround Remove the Celerra from the first domain before adding it to a different one. If you have already added it to both you can still remove it from the one that you are trying to get it out of.

File Type field is Blank in Export to File Dialog with JRE 1.7

The “File Type” field is empty when the “Export to File” dialog is displayed. This happens only with JRE 1.7. No file is created in the destination folder and the dialog closes without an error.

Workaround The workaround is to select the file type from the “File Type” drop down.

43



JRE install auto link gives an error

Unisphere will prompt a message to install JRE 1.6 if JRE is not installed. The JRE install link in this message does not work.

Workaround The workaround is to install JRE 1.6 directly from the Oracle website.

Alerts

For performance and memory usage reasons, the alerts displayed in Unisphere are now limited to 400. A new alert has been added to notify the administrator if there are more than 400 alerts:

Severity: Critical

System: <hostname>

Domain: Local

Created: <timestamp>

Message: Too many alerts on <hostname>.

Full Description: This system has more than 400 alerts. Until you delete the alerts, you will only see the top 400 alerts (by date and severity) for this system.

Recommended Action: Delete the alerts on this system to clear them from the list of alerts.

Event Code: 0x240060005

Do not attempt to delete the "Too many alerts" alert in Unisphere. Unlike most of the other alerts, this alert will clear itself if the condition is resolved and the total number of alerts for a system is reduced to fewer than 400. If you include this new alert in a selected list to be deleted, the alerts will appear to be deleted, but will be reloaded once a new browser session is started.

A failure will be posted to /nas/log/webui/apl_tm.log, but no failure will be displayed in Unisphere.

In /nas/log/webui/apl_tm.log:

2010-06-24 10:20:21.260

webui-N:[[email protected]]:0:29366:S:402345-Delete alert(s).

2010-06-24 10:20:28.783 webui-N::0:29366:E:402345-Delete

44



alert(s).(complete=true,failed=true) <- Note failed=true

Browsers

EMC requires the use of one of the following browsers to run Celerra Manager:

Internet Explorer 6.0 SP2, 6.2.3 through 7.x.; Internet Explorer 7.1 is recommended. Internet Explorer 8.0 is also supported.

Firefox 3.0 or later.

Firefox Firefox Password Manager (PM) is enabled by default. This feature records and stores the password that you enter on a website and then automatically fills in the password entry field on your next and subsequent visit. The password field masks the password with asterisks (*).

To prevent Firefox from remembering passwords:

1. Click Tools > Options... and select the Security tab.

2. Under Passwords, clear the Remember password for sites checkbox.

Browser session hang

Depending upon a client's particular browser configuration (browser type, version, JRE version, native OS, etc.), launching the USM tool from the UEM may result in a browser session becoming hung.

If you verify that there are no hidden dialogs awaiting confirmation, terminate the browser sessions and all associated processes. Rebooting the client system may be required to restore full browser operation. To prevent this issue in the future, launch the USM directly (i.e., not via UEM).

Changes from other user interfaces

Any changes made to the network server using another user interface such as the Celerra CLI are not automatically shown in Unisphere. Refresh the browser to see these changes.

Changing default password

Changing the CLARiiON array default admin password breaks the ability of the Celerra to communicate to the CLARiiON or make any changes to its database [diskmark, discoveries, etc. will all fail].

45



Checkpoint schedule page

If the time zone of the Celerra Control Station and the client system running the UEM are different the checkpoint schedule page will display times in UTC (coordinated universal time).

Confirmation dialog inactive

While using Unisphere, a confirmation dialog appears briefly, and then moves to the background. Users need to manually re-focus the window and bring the dialog box to the foreground.

Dashboard screen

The data is not presented in the Celerra Capacity View screen in an appropriate amount of time. However, the other dashboard screens are loaded successfully.

Workaround Close the Unisphere window and clean out the persistence files. On the next restart of Unisphere - you will not see this issue.

To clean out the persistence files, follow these steps:

1. Close the Unisphere application.

2. Go to the Unisphere client log folder; the default should be C:\Documents and Settings\<username>\emc\Unisphere.

3. Delete the persistence.ser file.

4. Re-launch the Unisphere application

Dashboard summary

The Dashboard Summary page did not contain view block. This is unexpected, but those blocks can be re-added using the customize button.

File system delete confirmation

The 'OK' button is disabled on the Unisphere File System Delete Confirmation page for a file system with File-Level Retention enabled at the Compliance level.

The table on the File System Delete Confirmation page indicates that protected files still exist on the file system even though the File System Properties page and the 'nas_fs -i' CLI command both show that all file protection has expired for the file system. The user is unable to delete the

46



file system using Unisphere.

Workaround:

The File-Level Retention status is now cached for performance reasons.

Force the file retention data to be updated by selecting the Unisphere Refresh for the File System List page (in the upper right corner of the File System List panel) or by opening a new browser session.

Help sessions

While getting a session timeout in a help session, some help text is displayed, but none of the icons are displayed. Clicking OK button has no effect. Workaround: Use the close button of browser to close the window.

JRE memory limit

For optimal UI performance, EMC recommends that the memory limit allocated to the JRE be increased. The default size of 64 MB is often too small and may result in Out of memory errors.

To set the memory size, from the Start menu in Windows, select Settings > Control Panel. Double-click the Java Plugin icon. Select the Advanced tab. In the Java Runtime Parameters, enter the following: -Xmx512m

The argument specifies the maximum memory. Typically, about ½ the size of your overall workstation's memory is recommended as the maximum memory value. For instance, if your workstation has 1 GB of memory, then use -Xmx512m for the JRE runtime parameter. If your workstation has 2 GB or more memory, you can adjust the max JRE memory as needed to avoid running out of Java memory and possibly improving overall performance.

Logging in after browser crash

After a browser crash, the GUI reports "system already managed" when user tries to log back in.

Logging in via multiple screens

Logging into Celerra via two or more screens may succeed, but logging out of one of the sessions causes the other sessions to lose contact.

Login screen

When you log into Unisphere via a Celerra system, Unisphere will not display a login screen if you click the “yes” button on the security

47



warning dialog, without also checking the “Always trust content” checkbox on the same dialog.

MMC snap-in

MMC installer program may not completely uninstall all components when requested to do so. However, when a new version is installed, the components are overridden and the application runs as expected.

Modifying system properties

When storage system properties (specifically - security credentials) are modified in the GUI, the following (example) warning message is displayed in addition to the expected icons: Modify CLARiiON CLARiiON CX4-960 APM00093300340. Invalid component information :BackendResume:

This error is a result of the component information cron job querying the status while the credentials are in the process of being modified. The correct component information will be refreshed during the next iteration of the cron job (approx. 5 to 10 minutes).

Mounts table view

The file system column in the Mounts table view gets stuck at "Loading" for multiple minutes. The rest of the table is populated correctly. Workaround: Refresh the page using the green arrow icon.

nbsnas/jserver filling up at 100%

Unisphere jserver mounted on /nbsnas/jserver filling up at 100% prevents installation. When this occurs, the following error is displayed:

The feature identified has detected a potential capacity issue that may need the attention of the system administrator. The device name and mount point is indicated in the event posting. Recommended Action: Check that the mount point's condition is in fact approaching cautionary levels.

Use normal administrative procedures to see if some free space can be returned to the mount point, such as removing unnecessary files or reorganizing areas of this mount point off onto other less full mount points on the system. If no means are available to reduce the space consumption on this mount point then you may want to investigate if there is a possibility that this mount point needs to be expanded to allow for more free space. Contact your Authorized Service Provider if you need assistance with this process or need to discuss acquiring additional storage system capacity.

48



Provisioning wizard

If an error occurs while in the Provisioning Wizard, you are instructed to correct the error and are given a choice of either restarting the provisioning process or canceling the activity. Rolling back intermediate tasks may fail or hang.

If you cancel the activity by clicking the NO button, all the previous tasks are rolled back and you will be redirected to the first screen of the Provisioning Wizard.

Slice volumes not listed in wizard

Occasionally, slice volumes created through the CLI or Unisphere Volume Create page are not listed as available volumes in the New File System Wizard when selecting Manual Volume Management and then attempting to create a new stripe volume from the slice volumes.

Storage domain

When the Celerra belongs to a storage domain that is destroyed, the storage domain users on the Celerra Control Station are not automatically moved to the "defunct" state. To correct this, manually remove all the storage domain users from the system by running the command:

/nas/http/webui/bin/admin_management.pl - defunctStorageUsers

Then remove the Celerra from the domain.

User’s tab displays unrelated messages

Unisphere > Settings > User's tab displays messages which are unrelated to that screen. These messages are a result of a query that was initiated by the user on some previous table screen.

Workaround Visiting another table screen and then to coming back to the table screen where the messages were seen should remove these messages.

Other restrictions

Caution Do not install Linux non-Celerra applications on the Control Station.

Doing so consumes resources and might adversely affect system operation. Celerra Control Station is a dedicated controller and not an applications platform.

49



Do not manually edit the nas_db database without consulting EMC Customer Service. Any changes to this file could cause problems with your Celerra installation.

Command line scripts

The NAS database stores specific information required for each Data Mover. The Celerra Network Server automatically performs a backup of the entire database every hour and saves it to a file named nasdb_backup.1.tar.gz. These backups start one minute after the hour.

EMC recommends that you check your scripts to make sure they are not scheduled to start at the same time as the backups. During backup, the database becomes locked or inaccessible, and some commands that rely on the database might fail.

Correctable hardware errors

Transient correctable hardware errors are currently being handled by the firmware. Although rare, if enough of these errors occur, a watchdog panic can be triggered. In a future release, these errors will be handled by the Celerra operating system instead of the firmware which will improve efficiency and prevent the watchdog panic.

server_cpu

The server_cpu ALL -halt now command does not complete when using Celerra NSX Network Servers. This is because the NSX needs at least one Data Mover to be up in order to access its NAS partitions through NBS.

server_http

The server_http command fails if you issue it on a Data Mover while running fsck. Other commands (including FileMover API) may not work until fsck completes.

server_stats

On a Data Mover with more than 500 mounted file systems, the following statistics fail to display: - store.logicalVolume.metaVolume - metaVolumes-std. If any of these statistics are supplied as an argument to server_stats -monitor option, the server_stats will hang. Use Ctrl-C to terminate the session.

server_sysconfig

If you use the option string of server_sysconfig, the entire string must be

50

Problem prevention


enclosed by double quotes.

Virtual device naming

When creating virtual devices, such as Fail-Safe Network (FSN) devices, Ethernet channels, and link aggregations, do not use the following names as the name of the device:

Amgr http Mac pfkey spstream tcprpc Blksrv i82596 meshforwarder pipe Ssl Tftp Client icmp nd pmap sslpipe Toeframechk Cmfap icmp6 nfs rcp statd upd Dep ip nfsPrimary rfa StreamHead udpnfsmesh dfe ip6 nwh scsicfg tcp

echo lockd pax smb tcpnfsmesh escon lpfn pcnfs smpte tcprpc

Problem prevention

The EMC Customer Service center has reported a number of calls concerning the following issues. Please take a moment to familiarize yourself with these potential issues.

Administrative activities

Log in as nasadmin

As with any UNIX-based operating system, refrain from performing normal administrative activities as root user. Log in using the administrative user account instead.

Shutting down a Data Mover or deleting an internal interface

Shutting down a Data Mover or deleting an internal interface will prevent you from issuing any server_ commands against that Data Mover. Whenever you issue any command on the Celerra Control Station CLI server, the command is sent to and executed by a Data Mover. If that Data Mover is down or unreachable, the command will fail.

51

Problem prevention


Celerra AntiVirus Agent (CAVA)

AntiVirus (AV) threads

If a large number of files need to be checked at the same time, the CAVA file checking process may be slow depending on number of threads configured, number of CAVA servers, and speed of the CAVA servers.

If the CAVA servers cannot process file checks quickly enough, then all AntiVirus (AV) threads will become busy and the AV “collector queue” will increase, indicating that file check requests are queueing up. Increasing the number of AV threads (10 by default) will allow more files to be checked in parallel and may improve performance. However, the number of CAVA engines may need to be increased to satisfy a heavy load.

Number of CAVA engines

It is possible to configure your EMC Celerra installation with only one Celerra AntiVirus Agent (CAVA) engine. However, this non-high-availability configuration can make your installation susceptible to data unavailability in the event that the CAVA engine goes offline for any reason.

The best method for preventing this type of outage is to increase the number of available CAVA engines.

Additional information for configuring the Celerra AntiVirus Agent can be found in the Using Celerra Anti Virus Agent technical module. This document describes the parameter setting that will specify shutdown actions to take when no servers are available.

Removing AV user from Data Mover password file

If you remove an AV user from the Data Mover password file, CAVA will not work. Depending on the setting you have on CIFS, you could use up all the CIFS threads because they are used for virus checking or it will shut down CIFS altogether.

CIFS

Deleting machine account from active directory

Deleting a machine account from active directory will result in loss of access to the CIFS server and its resources. Deleting the computer account of any CIFS server, including a Celerra Data Mover, from the domain of which it is a member will prevent that CIFS server from being

52

Problem prevention


able to authenticate both itself and users attempting to connect to it with the domain. This will mean that users will not be able to access data hosted by the CIFS server.

Deleting server out of the domain

Deleting a server out of the domain will result in shares associated with the server to be unavailable. The same applies to deleting the machine account.

Naming CIFS shares

Do not name your CIFS share "all" or you risk deleting all of your shares when trying to delete it.

NetBIOS name conflict

Instead of destroying the conflicting NetBIOS, simply rename it using the existing server_cifs <server_x> -rename -netbios <oldname> <newname> command. Changing a NetBIOS name using this command will update the share and local group and enable you to reuse it.

LGDB and ShareDB information associated with the server are persistent. If the target name of a rename conflicts with LGDB or ShareDB information, re-create the old server and rename it to a name that will not conflict with your naming structure to free the old slot.

Control Station

Applications

The Control Station’s primary function is to manage the Celerra. The use of applications that do not come installed on the Control Station is not recommended. The Celerra Control Station is not tested or certified with any of these applications and there could be issues with memory usage or CPU time that may affect normal operation.

Dual Control Stations

If you have dual Control Stations on either (CFS/CNS/NSxxx) and they concurrently try to access the /nas file systems, the nasdb may become corrupted and the Data Movers will be unable to communicate with the Control Station.

53

Problem prevention


File System

Deleting files from Data Mover or file system

Files that you delete will be permanently lost unless you have a backup copy. Files can be backed up and restored through the Celerra SnapSure, TimeFinder/FS, Replicator or by using standard tape backup methods.

Deleting file systems

File systems that you delete will be permanently lost unless you have a backup copy. Backups can be created via TimeFinder/FS, Replicator or other backup methods.

File system full

Insufficient space in a file system will prevent users from writing to that file system. In addition to monitoring file system usage, set an event notification for file system full condition and enable automatic file system extension with a conservative threshold.

Refer to the Configuring EMC Celerra Events and Notifications technical module for more information.

Mounting file system to /tmp

Some variants of UNIX or Linux periodically delete some or all of the files under the /tmp directory. If you mount a file system under /tmp, you risk having the contents of that file system deleted by the automatic cleanup process.

Splitting BCVs

When a production file system (PFS) is mounted in a read/write fashion to a Data Mover and has BCVs established, the command to split (or mirror -off) the BCVs from the PFS volumes must be initiated using the Celerra fs_timefinder command. The fs_timefinder command ensures consistency of the BCV snapshot by managing file system access and Data Mover log operations when the volumes are split. These additional considerations are not available directly from Solution Enabler commands (symmir).

Name Resolution

SunOne/iPlanet

When using SunOne/iPlanet, make sure that the defaultServerList

54

Problem prevention


attribute on the LDAP server is set to use IP addresses and not host names. If hostnames are specified and the Celerra is configured to use SunOne/iPlanet for host resolution, a deadlock can occur.

Networking

Changing an IP address

Changing an IP address on a Data Mover can result in a naming (DNS/NIS) mismatch and render the IP address inaccessible.

Changing or removing DNS settings

When you make changes to the DNS server, such as IP addresses make sure you reflect those changes in the Data Mover. Failure to do so may result in accessibility issues.

Removing the DNS server will prevent the CIFS server from communicating with the domain. Any CIFS server, including those configured on a Celerra Data Mover that is a member of a Windows 2000 or later domain, use DNS to locate the Windows domain controllers that it needs to communicate with to operate. Disabling DNS in such an environment will prevent the CIFS server from locating Windows domain controllers, therefore, prevent the CIFS server from authenticating users attempting to connect to it. This means that users will be unable to access the data hosted by the CIFS server.

Connecting an internal switch to an outside network

The internal networks in the Celerra are designed to provide private communication paths between Data Movers and the Control Station. If that network is connected to an external network, the resulting routing, addressing issues, and/or volume of network traffic may disrupt the communications between the Data Movers and the Control Station, which will in turn disrupt the operation of the Celerra.

Duplicate IP

Duplicating an IP address on two Data Mover interfaces will have the same result as duplicating IP addresses on any two machines in a network. The network will either be unable to determine which location it should deliver packets to or block all traffic to both machines on the assumption that one of them is attempting to gain unauthorized access to information by impersonating the other.

55

Problem prevention


FSN devices

Create Fail-Safe Network (FSN) devices on multiple network interface cards (NICs) to ensure connectivity in the event of a hardware failure.

Incorrectly adding network route

Incorrectly adding a network route will cause network access to be limited to the same subnet/broadcast domain. The default gateway, or route, address tells any network client, including a Celerra Data Mover or Control Station, where to send IP packets that are outside its local subnet. If the default gateway is misconfigured, the network client will not be able to contact any host that is not within its local subnet. If the Celerra relies on hosts, such as domain controllers, outside its local subnet, it may not be able to offer service.

Parameters

Changing parameters

Changing parameters could significantly affect the function of the Celerra Network Server. Take the following precautions before you make changes to Celerra parameters:

Do not change system parameters without first consulting with EMC Customer Service.

Change server parameters one at a time, in small increments, preferably in a non production environment. Be sure to thoroughly test each change before making additional changes.

Keep a detailed log of which server parameters you changed and the reason for the change. This information will be useful if a problem arises because of the change.

Managing parameters

EMC recommends using the server_param command or the Celerra Manager to manage parameters. If you manually edit the param file, note the following:

Parameters are case-sensitive and must be entered exactly as described in the Celerra Network Server Parameters Guide. A typographical error, the omission of a dot, incorrect capitalization, or a value out of range could result in adverse effects.

56

Technical notes


Storage

Multiple LUN 0s

Assigning multiple LUN 0s will result in Data Movers being unable to boot successfully.

Powering down all SAN Switches

The Celerra does not store any user data itself. All data is stored on the Symmetrix or CLARiiON disk array(s) that the Celerra is attached to. With a Celerra gateway model the Celerra is attached to the disk array(s) it uses via a Fibre Channel SAN. If that SAN is unavailable then the Celerra will not be able to access the disk array(s) it relies on and hence will not be able to offer normal service.

Reassigning volumes with SDDR or making fibre changes

Like any host of a Fibre Channel SAN or array, the Celerra will remember the address of the LUNs that it uses in the attached disk arrays. Any change in the identity or location of those LUNs without a corresponding reconfiguration of the Celerra may result in the Celerra being unable to access the LUNs it needs to provide normal service.

Reassigning Celerra LUN with Navi

Same as Reassigning volumes with SDDR or making Fibre changes.

Unbinding/rebinding LUNs

Same as Reassigning volumes with SDDR or making Fibre changes above.

Write disabling R1 and R2

In SRDF configurations, setting R2 disks to "read only" will cause R1 disks to become "read only" as well, and cause access issues.

Technical notes

Configuration guidelines

The values listed in this section represent the highest or lowest values tested by EMC. In some cases, actual capacities may exceed the capacities tested by EMC. Theoretical means in theory, not always practical. Refer to the E-Lab Interoperability Navigator for the most up-to-date information regarding these guidelines.

57

Technical notes


Table 4 CIFS Guidelines

Guideline/specification Maximum tested value

Comment

CIFS TCP connection 64K (default and theoretical max.), 20K (max.tested)

Param tcp.maxStreams sets the maximum number of TCP connections a Data Mover can have. Maximum value is 64K, or 65535. TCP connections (streams) are shared by other components and should be changed in monitored small increments.

Share name length 80 characters (Unicode). 12 chars (ASCII)

Unicode: The maximum length for a share name with Unicode enabled is 80 characters. ASCII: In ASCII mode, the maximum share length is 12 characters.

Number of CIFS shares 10,000 per Data Mover 40,000 per Celerra

You may notice a performance impact when managing a large number of shares.

Number of NetBIOS names/compnames per Data Mover

509 (max.)

Limited by the number of network interfaces available on the Data Mover. From a local group perspective, the number is limited to 509. NetBIOS and compnames must be associated with at least one unique network interface.

NetBIOS name length

15

NetBIOS names are limited to 15 characters (Microsoft limit) and cannot begin with an@ (at sign) or a - (dash) character. The name also cannot include white space, tab characters, or the following symbols: / \ : ; , = * + | [ ] ? < > “. If using compnames, the NetBIOS form of the name is assigned automatically and is derived from the first 15 characters of the <comp_name>.

58

Technical notes


Comment (ASCII chars) for NetBIOS name for server

256

Limited to 256 ASCII characters. Restricted Characters: You cannot use double quotation ("), semi-colon (;), accent (`), and comma (,) characters within the body of a comment. Attempting to use these special characters results in an error message. You can only use an exclamation point (!) if it is preceded by a single quotation mark (’). Default Comments: If you do not explicitly add a comment, the system adds a default comment of the form EMC-SNAS:T<x.x.x.x> where <x.x.x.x> is the version of the NAS software.

Compname length

63 bytes

For integration with Windows environment releases greater than Windows 2000, the CIFS server computer name length can be up to 21 characters when UTF-8 (3 bytes char) is used.

Number of domains

10 tested 509 (theoretical max.)

The maximum number of Windows domains a Data Mover can be a member of. To increase the max default value from 32, change param cifs.lsarpc.maxDomain. The Parameters Guide contains more detailed information about this parameter.

Block size negotiated

64 KB

Maximum buffer size that can be negotiated with Microsoft Windows clients. To increase the default value, change param cifs.W95BufSz, cifs.NTBufSz or cifs.W2KBufSz. The Parameters Guide contains more detailed information about these parameters.

Number of simultaneous requests per CIFS session (maxMpxCount)

127

This value is fixed and defines the number of requests a client is able to send to the Data Mover at the same time, (for example, a change notification request). To increase this value, change the maxMpxCount parameter. The Parameters Guide contains more detailed information about this parameter.

59

Technical notes


Total number files/directories opened per Data Mover

200,000 (NS/500/700, DM514, DM 514T,NSX, NS40, NS80)

A large number of open files could require high memory usage on the Data Mover and potentially lead to out-of-memory issues.

Number of Home Directories supported

20,000

Number of Windows/UNIX users

unlimited

Earlier versions of the Celerra Network Server relied on a basic database, nameDB, to maintain Usermapper and secmap mapping information. DBMS now replaces the basic database. This solves the inode consumption issue and provides better consistency and recoverability with the support of database transactions. It also provides better atomicity, isolation, and durability in database management.

Number of users per TCP connection

16K (default), 64K (max)

To increase the default value, change param cifs.listBlocks (default 64, max 255). The value of this parameter times 256 = max number of users. Note: TID/FID/UID shares this param and cannot be changed individually for each ID. Use caution when increasing this value as it could lead to an out-of-memory condition. Refer to the System Parameters Guide for parameter information

Number of CIFS threads

96 (1GB memory) 256 (>1 GB memory)

Number of threads for CIFS users (if there is 1 GB of memory on the Data Mover, the default is 96 threads; however, if there is over 1 GB of memory, the default number of threads is 256). This is the default, you can increase the number using the server_setup command with cifs start=xx value.

60

Technical notes


Number of CIFS connections (share accessed) per TCP connection


To increase the default value, change param cifs.listBlocks (default 64, max 255). The value of this parameter times 256 = max number of connection. Note: TID/FID/UID shares this param and cannot be changed individually for each ID. Use caution when increasing this value as it could lead to an out-of-memory condition. Refer to the System Parameters Guide for parameter information.

Number of files/directories opened per CIFS connection


To increase the default value, change param cifs.listBlocks (default 64, max 255). The value of this parameter times 256 = max number of files/directories opened per CIFS connection. Note: TID/FID/UID shares this param and cannot be changed individually for each ID. Use caution when increasing this value as it could lead to an out-of-memory condition. Be sure to follow the recommendation for total number of files/directories opened per Data Mover. Refer to the System Parameters Guide for parameter information.

Number of VDMs per Data Mover

29

Note: The total number of VDMs, file systems, and checkpoints across a whole cabinet cannot exceed 2048.

number of ntxmap User Mapping entries

1000

Table 5 FileMover


Comment

Max connections to secondary storage per primary (Celerra) file system

1024

Number of HTTP threads for servicing FileMover API requests per Data Mover

64

Number of threads available for recalling data from secondary storage is half the number of whichever is the lower of CIFS or NFS threads. 16 (default), can be increased using server_http command, max (tested) is 64.

61

Technical notes


Table 6 File systems guidelines Guideline/specification Maximum tested

value Comment

Mount point name length

255 bytes (ASCII)

The "/" is used when creating the mount point and is equal to one character. If exceeded, Error 4105: Server_x:path_name: invalid path specified is returned.

File system name length

240 bytes (ASCII) 19 chars display for list option

For nas_fs list, the name of a file system will be truncated if it is more than 19 characters. To display the full file system name, use the –info option with a file system ID (nas_fs -i id=<fsid>).

Filename length

255 bytes (NFS), 255 characters (CIFS)

With Unicode enabled in an NFS environment, the number of characters that can be stored depends on the client encoding type such as latin-1. For example: With Unicode enabled, a Japanese UTF-8 character may require three bytes.1 With Unicode enabled in a CIFS environment, the maximum number of characters is 255. For filenames shared between NFS and CIFS, CIFS allows 255 characters. NFS truncates these names when they are more than 255 bytes in UTF-8, and manages the file successfully.

Pathname length

1,024 bytes

Note: Make sure the final path length of restored files is less than 1024 bytes. For example, if a file is backed up which originally had path name of 900 bytes, and it is restoring to a path with 400 bytes, the final path length would be 1300 bytes and would not be restored.1

Directory name length

255 bytes.

This is a hard limit and is rejected on creation if over 255 limit. The limit is bytes for UNIX names, Unicode characters for CIFS. 1

Subdirectories (per parent directory)

65,533

This is a hard limit, code will prevent you from creating more than 65,533 directories.

Number of file systems per Celerra

4096 (NSX, NS40/NS80) 2048 (All other models)

This max number includes VDM and checkpoint file systems.

62

Technical notes


Number of file systems per Data Mover

2048

Mount operation will fail when the number of file systems reaches 2048 with an error indicating maximum number of file systems reached. This max number includes VDM and checkpoint file systems.

Maximum Disk Volume Size

dependent on RAID group (see comments)

Unified platforms — Running setup_clariion on the NS-960 platform will provision the storage to use 1 LUN per RAID group which may result in LUNS > 2 TB, depending on drive capacity and number of drives in the RAID group. On all CX4 integrated platforms (NS-120, NS-480, NS-960), the 2 TB LUN limitation has been lifted--this may or may not result in LUN size larger than 2 TB depending on RAID group size and drive capacity.For all other unified platforms, setup_clariion will continue to function as in the past, breaking large RAID groups into LUNS that are < 2 TB in size. Gateway systems — For gateway systems, users may configure LUNs greater than 2 TB up to 16 TB or max size of the RAID group, which ever is less. This is supported for NS-40G, NS-80G, NSX, and NS-G8 NAS gateways when attached to CX3, CX4 and Symmetrix DMX-4 backends. Multi-Path File Systems (MPFS) — MPFS supports LUNs greater than 2 TB. Windows 2000 and 32-bit Windows XP, however, cannot support large LUNs due to a Microsoft Windows OS limitation. All other MPFS Windows clients support LUN sizes greater than 2 TB if the 5.0.90.900 patch is applied. Use of these 32-bit Windows XP clients on NS-960 systems require RPQ (Request for Price Quotation) approval.

63

Technical notes


Total storage for a Data Mover (Fibre Channel Only)

Celerra Version 6.0 16 TB = NX4 32 TB =NS20, NS-120 64 TB =NS40, NS-480, NS40G, NS-G2 128 TB = NS80, NS80G, NSX 256 TB = NS960, NS-G8 Celerra Version 5.6 8 TB = CNS (514 DM) 10 TB = NX350, NS500 16 TB = NX4, NS700 32 TB = NS20, NS120 64 TB = NS40, NS480, NS40G, NS-G2, VG2 128 TB = NS80, NS960, NS80G, NSX, NS-G8

These total capacity values represent Fibre Channel disk maximum with no ATA drives. Fibre Channel capacity will change if ATA is used. Note: On a per-Data-Mover basis, the total size of all file systems, and the size of all SavVols used by SnapSure must be less than the total supported capacity. Exceeding these limit can cause out of memory panic. Refer to the Celerra Capacity Limits tables for more information, including mixed disk type configurations.

File system size

See right column for TB information per Data Mover

16 TB (hard limit, enforced) NS700, NSX, NX4, NS120, NS480, NS960, NS-G8 10 TB NS350, NS500 8 TB 514/514T Data Mover When planning a file system, with respect to maximum desired size, be sure to factor in the time required for FSCK, ACL checks, file system build & restore. This guideline applies to both Fibre Channel and ATA drives.

File size

16 TB (4 TB with file size policy of quota)

This hard limit is enforced and cannot be exceeded.

Number of directories supported per file system

Same as number of inodes in file system. Each 8 KB of space = 1 inode.

Number of files per Directory 500,000

Exceeding this number will cause performance problems.

64

Technical notes


Maximum number of files and directories per Celerra file system

256 Million (default)

This is the total number of files and directories that can be in a single Celerra file system. This number can be increased to 4 billion at file system creation time but should only be done after considering recovery and restore time and the total storage utilized per file. The actual maximum number in a given file system is dependent on a number of factors including the size of the file system.

1 All other industry-standard caveats, restrictions, policies, and best practices prevail. This includes, but is not limited to, FSCK times (now made faster through multi-threading), backup and restore times, number of objects per file system, snapshots, file system replication, VDM replication, iSCSI LUN replication, performance, availability, extend times, and layout policies. Proper planning and preparation should occur prior to implementing these guidelines.

Table 7 ISCSI guidelines Guideline/specification Maximum tested

value Comment

LUN size

Minimum: 3 MB Maximum: 2 TB less 1 MB

Because an iSCSI LUN requires overhead, you cannot create a LUN size equal to the total available space; the maximum size for a LUN is slightly less (typically 1 MB less) than the amount of available space in the file system.

LUNs per target

256

Each target can have as many as 256 LUNs (numbered 0 through 255). Windows, however, limits each target to 255 LUNs (numbered 0 through 254). To comply with the Windows limitation, Windows - compatible mode, which is enabled by default on the Data Mover, limits the number of LUNs to 255. If you use a non-Windows iSCSI initiator, you can disable Windows--compatible mode by modifying the iscsi.WinCompat parameter. Refer to the System Parameters Guide for parameter information.

LUN extension size

Minimum: 8 MB Maximum: same as LUN size

Due to a volume formatting limitation on Microsoft platforms, when extending a LUN, the total LUN size after extension must be greater than 129 MB.

65

Technical notes


Target alias

255 bytes

This name is an alias for the formal iSCSI name. The alias is an alphanumeric string of as many as 255 characters. The alias can include Unicode characters in UTF-8 encoding. Because the alias is the key field in the Celerra Network Server’s iSCSI databases, each alias must be unique on a given Data Mover.1

iSCSI qualified name (IQN)

223 characters

EMC strongly recommends you leave this field empty and let the Celerra Network Server generate a legal iSCSI name. If you type a name, be sure it is globally unique and in either IQN or EUI format. An IQN can have as many as 223 characters. The Celerra Network Server converts all uppercase characters in an IQN or EUI to lowercase characters. Only letters, numbers, "-", ".", and ":" are allowed in iSCSI IQN names.

CHAP secret

255 characters

CHAP secrets must be a minimum of 12 characters and a maximum of 255 characters. Microsoft iSCSI initiator supports CHAP secrets of 12 to 16 characters.

Snapshots per LUN 2000

Maximum includes temporary writable snapshots.

Replication session alias

64 characters

Replication application label 128 characters

Replication attached comment text or binary data

4096 bytes

Configured replication sessions per Data Mover

1024

Concurrent active replication sessions per Data Mover

256 (theoretical max)

Number of configured iSCSI replication sessions that can be actively transferring data at one time.

Destinations per source file system

4

Exceeding this number may cause performance implications on the source file system.

66

Technical notes


iSCSI timeout

See comments

EMC recommends client iSCSI initiator timeout be longer than failover time. Note: MS iSCSI Initiator has a default timeout of 1 minute.

Table 8 Naming Services guidelines


Comment

Number of DNS domains

3 - WebUI unlimited - CLI

Three DNS servers per Data Mover is the limit if using WebUI. There is no limit when using the CLI (command line interface).

Number of NIS Servers per Data Mover

10

You can configure up to 10 NIS servers in a single NIS domain on a Data Mover. Note: A Data Mover supports only one NIS domain. Each time you configure a NIS domain and specify the servers, it overwrites the previous configuration.

NIS record capacity

1004 bytes

A Data Mover can read 1004 bytes of data from a NIS record.

Number of DNS servers per DNS domain

3

Table 9 NFS guidelines


Comment

Number of NFS exports

2,048 per Data Mover Tested Unlimited theoretical max.

You may notice a performance impact when managing a large number of exports using Celerra Manager.

Number of concurrent NFS clients

64K with TCP (theoretical) Unlimited with UDP (theoretical)

Limited by TCP connections.

Netgroup line size

16383

The maximum line length that the Data Mover will accept in the local netgroup file on the Data Mover or the netgroup map in the NIS domain that the Data Mover is bound to.

Number of UNIX groups supported

64K

2 billion max value of any GID. The maximum number of GIDs is 64K, but an individual GID can have an ID in the range of 0- 2147483648

67

Technical notes


Table 10 Networking guidelines Guideline/specification Maximum tested

value Comment

Link aggregation/ether channel

8 ports (ether channel) 12 ports (link aggregation (LACP)

Ether channel: the number of ports used must be a power of 2 (2, 4, or 8). Link aggregation: any number of ports can be used. All ports must be the same speed. Mixing different NIC types (i.e., copper and fibre) is not recommended.

Number of VLANs supported 4094

IEEE standard.

Number of interfaces per Data Mover

45 tested

Theoretically 509.

Number of FTP connections

Theoretical value 64K

By default the value is (in theory) 0xFFFF, but it is also limited by the number of TCP streams that can be opened. To increase the default value, change param tcp.maxStreams (set to 0x00000800 by default). If you increase it to 64K before you start TCP, you will not be able to increase the number of FTP connections. Refer to the System Parameters Guide for parameter information.

Table 11 Quotas guidelines


Comment

Number of Tree Quotas 8191 Per file system Max size of Tree Quotas 256 TB Max number of unique groups

64K Per file system

Quota path length 1024

68

Technical notes


Table 12 Replicator V2 guidelines Guideline/specification Maximum tested

value Comment

Number of Replication sessions per Celerra

1365 (NSX) (other platforms): 682

Max number of Replication sessions per Data Mover

1024

This enforced limit includes all configured file system, VDM, and iSCSI replication sessions as well as copy sessions

Max number of local and remote file system and VDM replication sessions per Data Mover

682

Max number of loopback file system and VDM replication sessions per Data Mover

341

The max number of iSCSI LUN replication sessions per Data Mover is the difference between the max number of replication sessions (1024) minus the number of configured file systems and VDM replication sessions per Data Mover.

Table 13 SnapSure guidelines


Comment

Number of checkpoints per file system

96 read-only 16 writable

Up to 96 checkpoints per file system are supported, as well as 16 writable checkpoints.

Table 14 Virus Checking guidelines Guideline/specification Maximum tested

value Comment

Virus checker file pending 90% of available VNODES in system

The percentage of total cached open files available in the system that can be pending on Virus Checking. An event is sent to the Control Station when the maximum is reached. To change the default value, modify param viruschk.vnodeMax(default=2000) and/or param viruschk.vnodeHWM(default=0x5a(90%). Refer to the System Parameters Guide for parameter information.

On systems with Unicode enabled, a character may require between 1 and 3 bytes, depending on encoding type or character used. For example,

69

Technical notes


a Japanese character typically uses 3 bytes in UTF-8. ASCII characters require 1 byte.

Capacity guidelines Celerra capacity limits - NS series

Following is a list of the versions 6.0 Celerra Capacity limits for the NS series. The E-Lab Interoperability Navigator is updated monthly and contains the most up-to-date capacity limits.

Table 15 Celerra capacity limits – NS series NX4 NS20 NS120 NS40 NS480 NS80

(XB-60 or XB-65)

NS960

Usable IP Storage Capacity limit per blade (all disk types and uses)5

16TB4 32TB4,6 32TB6 64TB4 64TB 128TB4 256TB

Usable MPFS Storage Capacity limit per blade (all disk types and uses) 5

N/A 128TB1

128TB1

128TB1

128TB1

128TB1

256TB1

Max FS Size 16TB

16TB

16TB

16TB

16TB

16TB

16TB

Max # FS per DM/Blade 7

2048

2048

2048

2048

2048

2048

2048

Max # FS per cabinet 7

4096

4096

4096

4096

4096

4096

4096

Max configured replication sessions per DM/Blade for Replicator v2 3

1024

1024

1024

1024

1024

1024

1024

Max # of checkpoints per PFS 2

96

96

96

96

96

96

96

Max # of NDMP sessions per DM/Blade

4

4

4

4

4

4

4

70

Technical notes


Table 16 Celerra capacity limits – NS series (continued) NS40

G

NS-G2

NS80 (XB-60 or XB-65)

NSX (XB-60 or XB-65)

NS-G8

V-G2

V-G8

Usable IP Storage Capacity limit per blade (all disk types and uses)5

64TB4 64TB4 32TB4 128TB4 128TB 64TB 256TB

Usable MPFS Storage Capacity limit per blade (all disk types and uses) 5

128TB1

128TB1

128TB1

128TB1

128TB

128TB

256TB

Max FS Size 16TB

16TB

16TB

16TB

16TB

16TB

16TB

Max # FS per DM/Blade 7

2048

2048

2048

2048

2048

2048

2048

Max # FS per cabinet 7

4096

4096

4096

4096

4096

4096

4096

Max configured replication sessions per DM/Blade for Replicator v2 3

1024

1024

1024

1024

1024

1024

1024

Max # of checkpoints per PFS 2

96

96

96

96

96

96

96

Max # of NDMP sessions per DM/Blade

4

4

4

4

4

4

4

Notes:

1. No snaps or replication –If planning to use snaps or replication, the Data Mover capacity limit will be the same as non-MPFS.

2. PFS is Production File System.

3. A maximum of 256 concurrently transferring sessions are supported per DM.

71

Technical notes


4. When using Replicator –To achieve storage capacities listed, Replicator v2 is required.

5. This is the usable IP storage capacity per blade. For overall platform capacity, consider the type and size of disk drive and the usable Fibre Channel host capacity requirements, RAID group options and the total number of disks supported.

6. NS20, NS120 – All ATA configuration can go up to 48 TB per blade.

7. This count includes production file systems, user defined file system checkpoints, and two checkpoints for each replication session.

Celerra virus checking solution

The following table lists the requirements for using CAVA with third-party antivirus engines.

AV Vendor Celerra Version CAVA Version CA

eTrust r8.1 All currently-supported DART versions1

1.8.9; 2.2.4; 3.4.0; 3.5.6; 3.5.7.1, 3.5.8, 3.6.2 eTrust r8

TrendMicro ServerProtect for EMC NAS 5.3.1

All currently-supported DART versions1

1.8.9; 2.2.4; 3.4.0; 3.5.6; 3.5.7.1, 3.5.8, 3.6.2

ServerProtect for EMC NAS 5.5.8 Sophos AntiVirus 3.x All currently-supported

DART versions1 1.8.9; 2.2.4; 3.4.0; 3.5.6; 3.5.7.1, 3.5.8, 3.6.2

AntiVirus 5.x

AntiVirus 6.x

AntiVirus 7.x AntiVirus 9.x

Kaspersky

Kaspersky Anti-Virus for Windows Servers Enterprise Edition

5.6.46.4 and later Celerra Event Enabler (CEE) version 4.5.1

McAfee

VirusScan 8.0i All currently-supported 1.8.9; 2.2.4; 3.4.0; 3.5.6; 3.5.7.1, 3.5.8, 3.6.2

72

Technical notes


VirusScan 8.5i Patch 2 and above

DART versions1

Symantec

Symantec Endpoint Protection 11.04

5.6.48.x and later 4.5.2.2 and later

SAV for NAS 5.1.x All currently-supported DART versions1

3.6.2

SAV for NAS 4.3.x 3.5.7.1, 3.5.8, 3.6.2 1Currently supported NAS versions include the entire 5.6.x and 6.0.x families. Refer to the EMC E-LabTM Interoperability Navigator for the latest supported version information.

Celerra scripting guidelines

Following are guidelines to keep in mind when creating Celerra command scripts.

Issue Action Scheduling NAS Database Backups:

The Celerra Network Server backs up the NAS database that stores specific configuration information required for each Data Mover every hour, at one minute after the hour. During part of the backup, the database is locked, and some commands that rely on the database might not have access. EMC recommends that command scripts avoid starting at one minute after the hour. Note that scripts with complex commands that run for an extended period may overlap the backup period. The duration of the backup may vary. Use the following Linux command to check the state of the backup process prior to executing scripts: ps -ef | grep nasdb_backup. If a lock condition occurs, wait a few minutes and retry.

Command sequencing Some commands must lock the database in order to execute. If multiple user-entered commands or scripts are active at the same time, some of these commands may lock the database and prevent other commands from executing. To avoid this, you should sequence commands whenever possible.

Sleep statements Some processes within a script can take time to complete. Use proper timing and adequate sleep statements to prevent timing-related issues.

Pipe and grep

Piping script outputs through grep is a helpful tool to check the status of the script. Use periodic checks to grep for file or database locked messages, timeouts, resource unavailable warnings, and other failure or success messages, and use this information to check status, pause the script, or halt it.

Return code check

All commands return a UNIX-style return code (for example: 0 for success or 1 for failure) or a text-based status code (for

73

Technical notes


Issue Action example, done) that can be used to help determine if the command completed or if there was an error or a conflict with the NAS database backup, or other commands being run. If a lock condition occurs, wait a few minutes and retry. If you create and run scripts, be sure to incorporate return code checks and verify for proper return codes from individual operations.

Scripting examples: The RECOVERABLE variable contains the following errors to retry on:

Unable to acquire lock

Resource temporarily unavailable

Unable to connect to host

Socket: All ports in use

Database resource vanished

Connection timed out NAS_DB locked object is stale

Following is an example of what the RECOVERABLE variable looks like:

RECOVERABLE="unable to acquire lock|Resource temporarily unavailable|unable to connect to host|socket: All ports in use|database resource vanished|Connection timed out|NAS_DB locked object is stale".

The res variable contains the command output.

#!/bin/sh

########################################################

# File: nas_cmdrcvr.sh

# Created by: NAS Engineering

# Date: Thursday, May 25, 2006

# Version: 1.0

# Notes:

# 1) script will retry commands for specified period of time

#2) script will log messages to file only if there's

available disk space

########################################################

NAS_DB=/nas

export NAS_DB

PATH=$PATH:$NAS_DB/bin:$NAS_DB/sbin:/usr/sbin:.

74

Technical notes


export PATH

RETRIES=60

SLEEPTIME=30

LOGDIR="/home/nasadmin"

LOGFILE="$0.log"

LOGLAST="$0.last"

DISKFULL=98

RECOVERABLE="Resource temporarily unavailable|\

unable to acquire lock|\

unable to connect to host|\

socket: All ports in use|\

database resource vanished|\

Connection timed out|\

NAS_DB locked object is stale"

#

# function to log messages to a file

#

nas_log()

{

DISKCHK=`df -k $LOGDIR|awk 'NR>1{print $5}'|sed 's/\%//'`

# if there's enough free disk space, append to log

if [ $DISKCHK -lt $DISKFULL ]; then

TDSTAMP=`date +%Y-%m-%d' '%T`

echo $TDSTAMP: $LOGMSG >> $LOGDIR/$LOGFILE

fi

# regardless of available space, always write last error

echo $TDSTAMP: $LOGMSG > $LOGDIR/$LOGLAST

}

#

# function to execute (and potentially retry) commands

#

nas_cmd()

{

# initialize variable(s)

retry_count=0

75

Technical notes


# loop until either successful of retry count exceeded

while [ $retry_count -le $RETRIES ]; do

# execute command and gather response

RES=`$CMD 2>&1`

# check if response means command is recoverable

if [ `echo "$RES"|egrep -c "$RECOVERABLE"` -ne 0 ]; then

# check retry count

if [ $retry_count -ne $RETRIES ]; then

# retry count has not been exceeded

LOGMSG="Command ($CMD) failed with ($RES)...retrying in

$SLEEPTIME s"

nas_log

sleep $SLEEPTIME

else

# retry count has been exceeded

LOGMSG="Command ($CMD) failed with ($RES)...exiting (retry

count of $RETRIES exceeded)"

nas_log

exit 1

fi

else

# command was either successful or failed for an unknown

reason

LOGMSG="Command ($CMD) successful with ($RES)"

nas_log

retry_count=$RETRIES

exit 0

fi

#increment counter for retries

retry_count=`expr $retry_count + 1`

done

}

#

# main

76

Technical notes


#

CMD="nas_volume -d mtv1"

nas_cmd

Using the NAS database and query facility

EMC has partially changed the layout or format of Celerra internal databases. This change can impact the use of awk or grep utilities when used in scripts that assume specific positions of fields in databases.

To enable searching of the NAS database, Celerra has developed a new query subsystem that appears as a hidden option on some of the nas_commands. This query subsystem enables you to specify the information you are interested in, allows you to format the output, and is independent of the database format.

Caution Do not use grep and awk to scan the database files. Database positions may change and substrings may return false matches for database objects.

Following is an example of a query to view unused disks:

nas_disk -query:inuse==n -format:'%s\n' -Fields:Id

To filter root disks, refer to the List all non-root disks that are not in use in Table 10.

Examples Use the following commands to view the tags (fields) that you can query:

nas_disk –query:tags

nas_fs –query:tags

nas_volume –query:tags

nas_slice -query:tags

Table 17 contains a list of examples to help you get started. Note that these commands can be run on the Control Station CLI, so the hardcoded values can be replaced with shell script variables.

Table 17 Task and query examples

Task Example Query the ID of a named file system nas_fs -query:Name==RLL_fs10 -

format:'%s\n'-Fields:Id

77

Technical notes


Query the ID of a named file system without the new line

nas_fs -query:Name==RLL_fs10 -format:'%s'-Fields:Id

Query the name of a file system ID that corresponds to a particular ID

nas_fs -query:id==20 -format:'%s\n' -Fields:Name

List of all server ID nas_server -query:* -format:'%s\n' -Fields:Id

List of all server names nas_server -query:* -format:'%s\n' -Fields:Name

List of all the checkpoint file systems nas_fs -query:type==ckpt -fields:name -format:"%s\n"

List type of file system with ID 20 nas_fs -query:id==20 -format:'%s\n' -Fields:Type

List the file systems that are in use nas_fs -query:inuse==y - format:'%s\n' -Fields:Name or nas_fs -query:inuse==y -format:'%s\n' -Fields:Id

Identify file system of which file system ID 28 is a backup

nas_fs -query:id==28 -format:'%s' -Fields:BackupOf

List the name of the server with ID 2 nas_server -query:id==2 - format:'%s\n' -fields:name`"

View which volume file system is built on nas_fs -query:Name==my_fs -format:%d -fields:VolumeID

View the block count of meta volume nas_volume -query:Name==my_meta3 -format:%d -fields:Blocks

View the block size of meta volume nas_volume -query:Name==JAH_meta3 -format:%d -fields:BlockSize

Find which server IDs use fs123 nas_fs -query:name==fs123 -format:'%s\n'-fields:ServersNumeric

List all non-root disks that are not in use nas_disk - query:inuse==n:IsRoot==False -format:"%s\n" -fields:name

List unused volumes that contain “dc” in the volume name

nas_volume -query:inuse==n:IsRoot==False:name=dc -format:"%s\n" -fields:name

List all available disks on a particular storage device (symm_id is a script/env variable)

nas_disk -query:inuse==n:SymmID==$ symm_id:IsRoot==False -format:"%s\n" -fields:name

78

Environment and system requirements


Query operators

Table 18 lists the operators to use when building your queries:

Table 18 Operator definitions

Operator Definitions = Having the pattern (sub-string match)

== Exact string match

=- Integer Minimum (not less than)

=+ Integer Maximum (not more than)

=* Any

=^ NOT having the pattern

=^= NOT an exact match

=^- NOT Integer Minimum (is less than)

=^+ NOT Integer Maximum (is more than)

=^* NOT Any (None)

Environment and system requirements

Symmetrix microcode and CLARiiON FLARE levels

The E-Lab Interoperability Navigator contains the latest microcode and FLARE information. The E-Lab Interoperability Navigator is updated monthly and can be found on EMC Online Support at http://Support.EMC.com.

Centera FileArchiver not supported

The Centera® FileArchiver policy engine is not supported when using Celerra Network Server version 6.0.36.4 or later.

Java Runtime Environment (JRE)

Minimum JRE 1.6.

After February 2013, Oracle will no longer post updates of Java SE 6 to its public download sites. Existing Java SE 6 downloads already posted as of February 2013 will remain accessible in the Java Archive on Oracle Technology Network. This Unisphere Client will run with


79

Documentation


JRE 1.7 but it has not yet been fully qualified with JRE 1.7. Future releases of Unisphere will be qualified with JRE 1.7 and the future Unisphere Client can be used as a workaround if there are issues specific to JRE 1.7.

Documentation

Celerra Network Server Version 6.0 documentation

EMC provides the ability to create step-by-step planning, installation, and maintenance instructions tailored to your environment. To create VNX customized documentation, go to: https://mydocs.emc.com/VNX.

Documentation updates

CDMS

The Celerra Data Migration Service documentation leads the user through a multi-step process to execute a migration. The latest version of this document highlights the success and failure criteria for various steps in this process. Celerra CDMS Version 2.0 for NFS and CIFS provides more information about CDMS. The following have also been added as warnings and cautions:

A verify command should not be run if there are any migration threads running or active on other CDMS connections in the same file system.

A verify command should not be run if there is another verify command still in progress on a different file system on the same Data Mover.

Multibyte support for Celerra data elements

In the Celerra management interfaces, multibyte support allows to input, store, and display multibyte characters. Users can input and view Celerra object fields in their native language. The multibyte feature is supported since Celerra NAS 5.6 and later.

The following data elements are enabled with multibyte support and also support the Unicode 3.0 standard:

CIFS Share name

80

Documentation


CIFS share comment

CIFS export path

Tree quota pass

Tree quota comment Quota user name

CIFS server netbios name

CIFS server computer name

CIFS server alias

CIFS server comment

CIFS domain name CIFS workgroup

Organizational Unit (OU) name

DFS root name

CDMS source share name

CDMS local path

CDMS source export path Mount point name

File system name

Checkpoint name

NFS export name

NFS export path

DFS root name CDMS source share name

CDMS local path

CDMS source export path

Mount point name

File system name

Checkpoint name NFS export name

NFS export path

81




Celerra installation process

This section lists some facts you should know about the Celerra Network Server and the installation and configuration process before proceeding further.

How to prepare for Celerra installation

To prepare for Celerra installation and configuration, complete the following tasks:

1. Perform the pre-site installation tasks as required by EMC to complete the installation and connectivity of the Celerra Network Server. Tasks include preparing electrical facilities, wiring the network, adding a telephone line, checking environmental conditions and related requirements, as stated in the Professional Services Statement of Work (PSSOW).

Note: For a copy of the PS-SOW document, contact your Professional Services Account Team. For examples of Celerra network topologies, refer to the Configuring and Managing Celerra Networking technical module.

2. Configure your environment for UNIX clients, Microsoft Windows clients, or a combination of Windows and UNIX clients, as applicable.

3. Complete network assessment to ensure proper performance and protocol configurations. (Optional EMC services can be ordered to assist with this process.)

Celerra Network Server installation and configuration

EMC Customer Service installs the Celerra Network Server hardware—a process that includes the physical setup, cabling, power application and diagnostic testing of the Celerra Network Server.

Next, EMC Professional Services or an EMC partner might configure and integrate the Celerra Network Server. For example, for a Celerra Network Server with a Symmetrix system, this process typically includes the following tasks:

Coordinating the activities to enable installation, cabling, and connectivity to system cabinets

Loading and configuring the Celerra Network Server software

82



Creating, mounting, and exporting file systems

Configuring network protocols

Integrating with a supported, existing user authentication environment

Configuring standby Data Mover(s), if applicable

Configuring and demonstrating system management and administration tools

EMC Professional Services can perform additional Celerra Network Server configuration tasks, such as host provisioning (with Celerra MPFS), backup protection, and network connectivity, depending on the individual EMC customer agreement defined in the PSSOW. These services are part of the EMC Network-Attached Storage Design and Implementation service offering, which is designed to provide continuous technical expertise as you manage your organization’s growing network storage needs. For more information about this service offering, contact your EMC Customer Support representative.

After EMC installs, configures, and tests the Celerra Network Server, it is fully operational and ready to support networked storage operations.

Celerra upgrades

You can upgrade the Celerra NAS software two ways:


Command Line Interface


You can upgrade the software for many Celerra integrated systems by using the Celerra Upgrade Tool (CUT), which is part of the Unisphere Service Manager (USM). CUT lets you upgrade Celerra systems without assistance from EMC or Customer Support by communicating over the Internet to EMC Online Support, and over a local network to a Celerra system. Three wizards guide you, step-by-step, through the process of upgrading the Celerra software, including updating the Control Station and the blades, but not the storage system:

Go to EMC Online Support at http://Support.EMC.com for information about USM and CUT and for information about downloading a copy of USM to a Windows system. The Upgrading Unisphere for File online help on USM provides information and instructions about using CUT and its wizards.

The Celerra Upgrade Tool can upgrade Celerra Network Servers running software versions:


83



• 5.6 to 6.0

• 6.0 to 6.0.x or later

The Celerra Upgrade Tool cannot upgrade systems with dual Control Stations, nor can it be used to upgrade Gateway systems. If your Celerra is not eligible for upgrade by means of CUT, you will need to use the CLI to upgrade your system.

Running multiple upgrade wizards The USM Celerra Upgrade Tool wizard does not explicitly prevent CLARiiON upgrade wizards from running while it is in use. The Celerra Upgrade Tool presents a prominent warning on its 'Welcome' page in the GUI. Note that this scenario is the same as simultaneously running multiple instances of USM and/or CLI against the systems, which cannot currently be prevented. Also, the CLARiiON wizards correctly prohibit the Celerra Upgrade Tool from launching while they are in use. In general, upgrading the Celerra NAS and the CLARiiON FLARE software at the same time can cause serious system problems and result in system downtime.

Other CUT restrictions:

Selecting and rebooting blades at different times can cause errors. Select and reboot all blades at same time during the upgrade in the Reboot Blades dialog.

The Celerra Upgrade Tool cannot upgrade systems with dual Control Stations.

NAS and FLARE upgrades

EMC requires NAS upgrades to be performed before FLARE is upgraded on any attached CLARiiON.

The reason for this is to maintain compatibility between the storage API (NAVI) on the Control Station and the FLARE on the CLARiiON. The API is built in as part of the NAS management software and is used to gather data about the backend such as statistics, health, and configuration.

If the FLARE is upgraded first, the API may encounter issues while communicating with the new FLARE. Upgrading the NAS first allows the API to have limited backward compatibility with the FLARE code.

Out-of-family upgrades

The following issues have been reported during out-of-family upgrades.

84



Error message during upgrade While upgrading from version 5.6 to 6.0, the following error message is displayed: Error - cannot open /usr/sbin/ext2online /dev/hda1. This error can be ignored, as it does not stop the upgrade and the task completes successfully.

Pre-Upgrade Health Check (PUHC) fails When attempting to upgrade the software via CLI from version 5.5 to 6.0, PUHC fails abruptly with a run-time failure message which does not explain the cause of the failure. Note: A 5.5 to 6.0 upgrade is not supported.

Upgrade Replication task While performing an out-of-family upgrade (5.6.x to 6.0.x) on a dual Control Station system, the Upgrade Replication task exits with the message:

=======================Tasks=======================

23:06 [ 3/80 ] Upgrade replication

0 seconds

Please shutdown control station in slot 1 and retry

When this occurs, users need to power off the secondary Control Station, just running 'shutdown -h' is not enough. The secondary Control Station must be in 'power-off' mode during the upgrade.

Reboot required at upgrade

CNS software The CNS software is comprised of two installation components, a Linux package and a CNS software package. The software currently running on the Celerra and the CNS upgrade software versions determine the number of Control Station and Data Mover reboots required and whether Data Mover reboots can be scheduled to a later time after the upgrade is complete.

The DART software runs on the Data Movers and is part of the CNS software package. The new version of DART is loaded and run on the Data Movers only after a Data Mover boot.

Control Station rebooting Most software upgrades require a Control Station reboot. When upgrading CNS software, if either of the first two digits of the new version number change (for example, 5.6.x to 6.0.x), the Control Station requires rebooting.

85



At the beginning of upgrade, the system information section lists whether or not the Control Station reboot is mandatory. Note that this initial output is long and may scroll off the screen. You will be prompted whether to continue the upgrade after all the upgrade information is displayed.

If you enter no, the upgrade stops, and you are returned to the command prompt.

===========System Information===========

Install Manager Version: 6.0.34-4

Install Manager Command:

/celerra/upgrade/pkg_6.0.34.4/install_mgr

Starts at: Thu Jun 3 10:36:56 2010

From version: 5.6.49-1

To: 6.0.34-4

Dual Control Stations: yes

Slot: slot_0

Blade slot(s): 2, 3, 4, 5

Cabinet family: NS

Cabinet type: NS-HAMMERHEAD

Control Station reboot: multiple (mandatory)

Blade reboot: 1 time(mandatory)

Total number of attempts: 0

=========================================

.

.< other information is posted here>

.

Do you wish to continue with the upgrade now [yes or no]? >

yes

Data Mover rebooting The Data Movers require a reboot in order to run the new version of DART. When upgrading CNS software, if either of the first two digits of the new version number change (i.e., 5.6.x to 6.0.x), the Data Mover requires rebooting.

86



The system information screen is the same as the information in the Control Station rebooting screen.

Limited scheduling of the Data Mover reboot exists. An estimated time for when the Data Mover reboot will happen is shown at the beginning of upgrade:

==============Estimated Time for Services ===============

Current Time: 10:36

Estimated Time when NAS service will be stopped: 10:38

Estimated Time when Data Movers will be reset: 11:28

Estimated Time when NAS service will be restarted: 11:39

Estimated Time when upgrade will be complete: 11:50

==================================================

The Upgrade Data Movers task prompts before it reboots the Standby Data Movers. It then provides a menu to reboot the Primary Data Movers similar to the following:

==================================================

11:58 [ 46/70 ] Upgrade Data Movers 7 minutes

Start at: Thu Jun 3 11:58:38 2010

All servers must now be rebooted.

Ready to reboot the Standby Data Mover (3). Press the ENTER

key to begin...

The following server(s) need to be rebooted to run the new

release:

2) server_2

Enter slot number(s) or ALL to select server(s) for reboot:

==================================================

When upgrading DART, if both of the first two digits of the new version of software remain the same (for example, 6.0..x to 6.0.y), you may delay the rebooting of the Data Movers.

The system information screen is the same as the information in the Control Station rebooting example except for the following: Data Mover reboot: 1 time (advisory)

The Upgrade Data Movers task asks if you wish to reboot the Data Movers. If you enter yes, you will be prompted to press ENTER to reboot the standby Data Movers, then prompted to reboot the primary Data

87



Movers. If you enter no, the Data Movers continue running the current (pre-upgrade) version of CNS software until they are rebooted.

=================================================

12:29 [ 44/67 ] Upgrade Data Movers 7 minutes

All servers need to be rebooted to run the new release.

Do you wish to reboot them now? [yes or no]:

Ready to reboot the Standby Data Mover (3). Press the ENTER

key to begin...

The following server(s) need to be rebooted to run the new

release:

2 ) server_2

Enter slot number(s) or ALL to select server(s) for reboot:

Waiting for ping response from server_2... :up (0 secs)

=================================================

Schedules during upgrades

As part of the upgrade procedure, schedules that run between midnight and 1:00 A.M. may lose those run times. This affects only schedules that run every X hours of the day, (X being from 1 through 12), and not schedules that run at specific hours of the day. This is a known problem and can be corrected by modifying the affected schedules after the upgrade.

Upgrade requirements

Existing read/write-able file system must have 128K minimum free space prior to upgrading to CNS version 6.0.

CLARiiON upgrades

Occasionally, CLARiiON storage connected to a Celerra Network Server requires a software upgrade. When upgrading CLARiiON software, do not run heavy loads on the Celerra, as operations may be extremely slow or interrupted.

Notes CLARiiON array software cannot be upgraded unless the Celerra

Control Station software is upgraded first. Ignoring this rule may make the CLARiiON unmanageable by the Celerra.

Incremental CLARiiON arrays cannot be added to systems running

88

Fixed problems


older Celerra software unless they are running a compatible release of CLARiiON software.

A warning will appear when you upgrade the Celerra software if the CLARiiON software needs to be updated. If you receive this warning, finish upgrading the Celerra software and then upgrade the CLARiiON software as soon as possible to avoid compatibility issues.

Consult the E-LabTM Interoperability Navigator at http://Support.EMC.com for all software interdependencies.

Fixed problems

Fixed problems categories

The Fixed problems section is grouped by the following categories:

Category Meaning

CEE Celerra Event Enabler, CAVA

Celerra Management Applications

Unisphere, Celerra Startup Assistant, MMC snap-ins, and so on.

CIFS Common Internet File System (Quotas, Usermapper, ans so on)

Control Station Celerra Control Station

CORE Celerra Data Movers

Deduplication Deduplication

FileMover Celerra FileMover

FLR File-Level Retention

Hardware Hardware

MPFS Multi-Path File System

Install, Config, and Upgrades Celerra installs, callhomes, configuration, upgrades Internationalization Internationalization (I18N)

NDMP/Backup Network Data Management Protocol or Backup

Networking Networking issues

NFS Network File System

Replication Celerra Replicator, SnapSure, Checkpoint

Security LDAP


89

Version 6.0.70.4


SRDF Symmetrix Remote Data Facility Storage

Storage Storage

Tools EMCOPY, EMCACL, LGDUP, SHAREDUP

UFS UNIX File System

Version 6.0.70.4

Backup

Severity Symptom description Fix summary Tracking number

1

Data Mover panic Page Fault Interrupt caused by sending NDMP an empty software set file filter or directory filter string.

The code has been modified to fix the filter logic for the empty file filter string or the directory filter string.

496542 / 48722540



2

High CPU and memory usage by the XML API server process when running the VNX Monitoring and Reporting Tool. All CS operations were impacted.

The code has been modified to resolve the issue by optimizing the Tree Quota query processing on the XML API server.

455143 / 43142954

1

On file systems with a very large number of acls, running the acl database printstats command might take a very long time.

The code has been modified to avoid collecting the wasted space statistics. The former behavior is still available by running acl database printstats slow.

467244 / 45564250

90

Version 6.0.70.4


CIFS


2

Could not immediately connect to share on the system when many users were connecting and disconnecting simultaneously. Sometimes could not connect to share when the system was overloaded by share connections.

The code has been modified to improve the connection logic.

495620 / 48114692

1 Data Mover panic caused by receiving an invalid Kerberos service ticket.

The length of the encrypted ticket component is now checked before allocating memory, and an error will be returned if the length is zero. Also, the share mount request fails, and the error is logged in the server log by the calling code.

505036 / 49707766

1

When all Common Event Publishing Agent servers were offline, the CIFS access could be lost briefly if the fault-tolerance log file became full.

The single sparse FT log file has been replaced by two regular files which will be written to one by one. When the first is full, it will be renamed with a bak extension and the second one will be created.

512489 / 50330376

2

Poor CIFS performance leading to DU could be caused by unnecessary virus scans after a file was opened as read-only.

The code has been changed to not trigger the scan if a user closes a file opened as read-only, after another user has previously opened and modified it.

519560 / 50672554

1

Issues were encountered after moving a compname to another VDM due to the Active Directory replication problem, especially for sites with multiple Domain Controllers.

The code has been modified to move a joined compname tranparently accross VDMs of the same Data Mover.

522085 / 51136100

91

Version 6.0.70.4


CIFS Virus Check


2

When the AV Engine was not available, the following message was logged: Warning: RecvStream failed I/O error.

The code has been modifed to suppress alert messages that occur while killing a MSRPC client connection.

499970 / 48272012

2

CEPP events were generated even if the user had the CEPP bypass right. CAVA/CEPA users with "bypass" rights can be suppressed.

The code has been modified to disable CEPP events for users who have the CEPP bypass right, and also to enable the CEPP event bypass right for users with this right who are not on a CAVA/CEPP server or connected on CHECK$.

504473

2

When a file system was unmounted with files waiting to be scanned by a virus checker, the files were not scanned after the remount of the file system on another path.

The code has been modified to recognize this state and handle it correctly.

506486 / 49683544

Control Station


2 The checkpoint creation on the VDM root file system could fail during the VBB backup.

The code has been modified to mount NDMP checkpoints of a VDM root file system to the VDM server.

476488 / 46629630

2 The Data Mover could fail to start fsck on a writeable checkpoint mounted on an unloaded VDM.

The code has been modifed to send the fsck request to the parent Data Mover.

502351 / 49425182

92

Version 6.0.70.4


2 A CIFS server created from GUI could fail to join the domain due to a leading dash in the administrator's password.

The code has been modified to correctly handle such a password.

503070 / 47564604

2

NDMP backup failed for the production file system mounted with the mount point that contained two root vdm prefixes, for example, /root_vdm_1/root_vdm_1/productin_fs.

The code has been modifed to avoid this problem.

525208 / 51439440

2 The SNMP trap did not get sent for a file system notification.

The code has been modified to avoid this problem.

487350 / 47698828

DART Core


1

An idle Data Mover might panic with SYSTEM WATCHDOG or System clock was stalled when there was no explanation for why this happened.

The code has been modified to better handle the case when clock interrupts stop unexpectedly and to allow more time for the system to recover.

500581 / 49117864

EE Escalations


2

If the Control Station couldn’t resolve the IP address of ftp.emc.com, the following error could appear in the system log: transfer_support_materials[24118]: The transfer script has failed. It cannot transfer data to ftp.emc.com or the remote directory. Please contact your authorized service provider to make sure the remote server has sufficient free disk space (100MB).

The code has been modified to use the new IP address for ftp.emc.com as the default IP address for transferring support materials.

520928 / 48169232

ftp://ftp.emc.com/

ftp://ftp.emc.com/

93

Version 6.0.70.4


FLR


1

Data Mover panicked due to the use of a NULL pointer when a file system that is mounted read-only attempted to log something to the flrLogHandler.

The code has been modified to check for read-only file systems before they attempt to use the flrLogHandler.

505349 / 49743518

IP Replication v2


1

The memory leak in the Replication V2 error path was caused by the improper manual cleanup of a session. This could result in an out of memory panic.

The code has been modified to remove the memory leak that occurred during some error recovery operations.

494281 / 48485034

NFSv4


2

The NFSV4 access might take a long time if multiple clients work on the same files. The delegation recalling process led to poor performance, especially when some clients did not reply to delegation recalls immediately.

The code has been modified to enhance the locking mechanism, allowing multiple requests to run in parallel.

490956 / 48099274

1 Page Fault Interrupt. Virt ADDRESS: 0x62f870 Err code: 0 Target addr: 0x10

The code has been modified to synchronize the release of a lock owner with other lock owner operations.

492876 / 48323880

1 The NFS access was lost during backup.

The code has been modified to avoid a deadlock related to the NFSv4 delegation recall and conflict checking.

496917 / 48792858

94

Version 6.0.70.4


1

When using NFSv4, Data Mover panicked with a dump header similar to: DART panic/fault message: *** GP exception. Virt ADDRESS: 0x00006af142. Err code: 0 ***

The code was modified so that the NFSv4_setAttr could write, exceeding the mblk limit, in this rare condition.

502522 / 49446140

1 The NFSv4 service was not responding; "Pool:NFSD_Exec BLOCKED" appeared in the datamover log.

The code has been modified to resolve the dead-lock issue.

503276

2 The NFSv4 response was slow with the LDAP resolver.

The code has been modified to no longer query group members.

525643 / 50177872

Quotas


1

When the user renamed a directory on which online quota check was running, a race condition could occur which caused threads to be blocked and file system to be inaccessible.

The code has been modified to eliminate this race condition.

509083 / 50156424

1 The system would panic if users tried to enable quota on an MGFS filesystem.

The code has been modified to disallow the quota feature on MGFS to avoid the panic.

519520 / 50889786

RDE - PE


2

The state of deduplication showed 0% saved data even though the files had been deduplicated. This occured when a deduplication scan was not completed due to a file system or a checkpoint resource error.

The code has been modified to allow the F-RDE PE to continue scanning the file system after encountering a resource error.

372977 / 35129780

1

The hash table mechanism used by the Policy Engine to track hard links could cause memory fragmentation issues after the table was deleted.This could lead to an out of memory panic with a header like: PANIC in file: ../bufpool.cxx at line: 1309 : couldn't get a free page or PANIC in file:

The code has been modified to create a private hash table using a memory mechanism that avoids the risk of fragmentation.

499671 / 49145068

95

Version 6.0.70.4


../addrspac.cxx at line: 503 : Out of memory

1

The Data Mover became inaccessible after the EM copy was done. Need to reboot the Data Mover to get access back.

The code has been modified to handle dedupplication of empty files in a manner that does not lead to poor performance.

512668 / 50460864

Transport Protocols


1

Under rare circumstances, Data Mover might panic with the following stack: gdb) bt #0 PANIC (string=0x1918be0 ">>PANIC in file: ../malloc.cxx at line: 662 : \n Memory corrupted") at ../panic.cxx:277 #1 0x0000000000243e60 in free (base=<value optimized out>) at ../malloc.cxx:662 #2 0x00000000002440e2 in operator delete (p=0x500) at ../_delete.cxx:23 #3 0x0000000000485def in CLIENT::operator delete (p=0x500) at ../client.cxx:522

The code has been modified to remove the real-time race condition by adding synchronization in deleting paths.

473726 / 46291616

UFS


1

The Data Mover used the netcap command to capture the network traffic of a network device and store the network packets to an internal file. When the file was full, a staging file would be used. If the staging file was not successfully opened, the Data Mover would panic.

The code has been modified to check the file handle of staging buffer before it is used.

511398 / 50349810

96

Version 6.0.70.4


1

When the Data Mover was enumerating file system checkpoints from a Windows client, it panicked with the following DART panic/fault message: Page Fault Interrupt. Virt ADDRESS: 0x000191688f Err code: 0 Target addr: 0x0000000008 **. This happened when at least two checkpoints for the file system had different time stamp for Month field while having the same stamp of Seconds, Minutes, Hours and Days.

The code has been modified to correct an invalid boundary check. This resolves the issue of the incorrect return of a NULL pointer for a valid date range.

523621 / 51266466

UFS Log


1 Overflow of an array in a filesystem in-memory data structure led to a deadlock and hang condition.

The code has been modified to increase the size of the array, to sense and handle the overflow, and to track current users of the array elements to enhance diagnosability.

522045 / 50805196

Backend Features


1

"nas_rdf -restore" failed with storage profile mismatch error if the file system was created from a storage pool.

The issue is resolved by refreshing storage profile name reference for rdf restore.

531824

System Management


2 Hitachi EFD was showed as STD disk in the gateway system.

This issue is fixed by adding the code to correctly recognize the drive type from Hitachi.

49746724/508000

97

Version 6.0.70.4


2

VBB backup of VDM root file system failed.

This issue has been fixed by mounting NDMP checkpoints of VDM root file system to VDM server instead of server_2.

46629630/476488

1

XML API server processing cost high CPU and memory usage when running VNX Monitoring and Reporting Tool. All CS operations were impacted .

The issue has been resolved by optimizing the Tree Quota query processing on XML API server.

43142954/455143 51087704/521751 50912918/522374

2

SNMP trap did not get sent for a filesystem notification.

The issue has been resolved by improving the code responsible for configuration of SNMP traps.

47698828/487350

2

It failed to start fsck on a writeable checkpoint mounted on an unloaded VDM.

The issue has been resolved by sending fsck requests to the parent Data Mover of the unloaded VDM.

49425182/502351

3

CIFS server created from GUI failed to join the domain due to a leading dash in administrator's password.

The issue has been resolved by adding a transient internal password prefix.

47564604/503070

1

NDMP backup failed for production filesystem mounted with the mount point containing two root VDM prefixes, for example: /root_vdm_1/root_vdm_1/productin_fs.

The issue has been resolved by removing the code that strips off the root VDM prefixes from the mountpoint.

51439440/525208

98

Version 6.0.65.2


2

NDMP backup failed due to automatic ckpt creation failure when the file system was created with a physical DMbut mounted with a VDM path through GUI.

This issue has been fixed by adding a check that will prevent the GUI from creating a file system with a physical DM but mounted with a VDM path.

48710464/496443

Version 6.0.65.2

Backup

Severity Symptom Description Fix Summary Tracking Number

1 The Data Mover panicked with a Page Fault Interrupt when VBB backup ran on a file system larger than 2TB.

The code has been modified to prevent panic due to integer overflow.

458269 / 44465210


Severity

Symptom Description

Fix Summary

Tracking Number

1 Users could not log on to Unisphere if the passwords contained URL special characters like "&, ; ,+".

The issue has been fixed by encoding the password and user name to URL encoding.

396490 / 38065374

2 "% CPU Idle" graph on "Resource Utilization History" page in Celerra Monitor showed value of 0%.

This issue was fixed by correcting the formula for calculation of the graph data.

481652 / 46960472

CIFS


2 Files renamed by SMB2 clients could not be submitted to the virus checker.

Code has been added to submit files renamed by SMB2

472623 / 45655462

99

Version 6.0.65.2



clients to the virus checker.

2

The user was not able to delete a DFS link that existed beneath another DFS link.

The code disallows creation of a DFS link beneath another DFS link. This is consistent with Windows server behavior.

475820 / 46376654

1

When a user changed the UNIX rights of a file located on a file system mounted with MIXED access policy, the ACE displayed on Windows showed only "Owner Rights". To check who corresponded to this ACE, the user should go to the owner tab. The access checking was not affected.

The ACE now displays the owner name resolved through NIS and its uid number. It is no longer necessary to go to the owner tab.

478365 / 46563750

2

Quotas enumeration stopped responding indefinitely on Windows7 and W2K8 clients (clients using SMB2 protocol).

The code has been fixed to restore ability to enumerate and report user' s quota.

481974 / 46640214

2

A file could not be created on File Level Retention (FLR) if scan on read was enabled.

The code has been updated to not scan a file if it has just been created during the open operation.

482464 / 47200498

1

server_cifssuport -cred -path <path> command panicked on Data Mover when the path specified was longer than 220 characters and was unreachable.

The code has been fixed to correctly handle unreachable path even when it is longer than 220 characters.

486226 / 47626966

1

Data Mover CPU consumption reached 100% when there were a lot of file range locks.

The code has been improved to rearrange some tests in the lock manager so that checking lock conflicts consumes less CPU.

487750 / 47801326

2

Internet Explorer returned an error when downloading files on Windows 7 with 64bit CAVA.

The code has been fixed to allow open a file for read attribute even when another process has the file open.

490177 / 47973450

100

Version 6.0.65.2



2

OS X 10.7 could not mount CIFS FS because the reply of opening CIFS might report invalid maximum access rights of the logged user if rights were set from NFS on a file system mounted with mixed access policy.

The open reply now correctly reports maximum access rights of the logged user.

491591 / 45540912

1

The Data Mover panicked when virus checker was enabled with param waitTimeout set. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x000110f214 Err code: 0 Target addr: 0x0000000800 **

The code has been fixed to avoid the panic condition.

496627 / 48742734

1

Data Mover panicked with “out of memory” when CEPP was started with FT level=1 and no more CEPP servers were available. The following message was displayed: DART panic/fault message: >>PANIC in file: ../bufpool.cxx at line: 425 : couldn't get a free page


499427 / 49063234

CIFS Virus Check


2 The disconnection of a client did not unblock CIFS threads.

The code has been fixed so that when a client disconnects, all CIFS threads working for the client are unblocked.

476743 / 46658574

2 CEPP generated a “file open read” event incorrectly instead of “open dir” event.

The code has been fixed to generate the correct event.

500020

101

Version 6.0.65.2


Control Station

Severity

Symptom Description

Fix Summary

Tracking Number

1 Some of the Unisphere GUI pages like "New File System Page" and "File System Properties Page" took over 20 minutes to load on a gateway system connected to Symmetrix array.

The performance issues have been fixed by revising the code. The Unisphere GUI pages take approximately 2 seconds to load with the fix.

467458 / 45558122

2 Creation of a Virtual Data Mover on a mirrored storage pool failed with an unclear error message "No space left in matching storage pools".

CCMD message definition has been modified to clarify the term "matching storage pools".

375163 / 35536804

2 nas_quotas report displayed "EXPIRED" in "Timeleft" column even when filesystem quota grace period was defined as "unlimited".

This issue has been fixed by displaying "NoLimit" instead of "EXPIRED" when file system quota grace period is defined as "unlimited".

466085

2 nas_diskmark failed to mark a new LUN and displayed a warning similar to the following:- "Warning: 17716815787: server_2 c16t0l6 skipping reserved LUN id, APM00111803762 stor_dev=0x00ED (237 decimal), maps to a reserved host LUN id 0x6."

This issue has been fixed by removing an old restriction which limited the new HLU number to be assigned to a newly discovered LUN.

472963 / 46066538

2 If nas_diskmark executed on a gateway system tried to attach to a Symmetrix array, it failed with an error similar to the following:- "Error 5008: server_5 :13421849340: File system root_avm_fs_group_1 is made up of disk volumes with inconsistent density."

This issue has been fixed by limiting filesystem inconsistent density check to CLARiiON-based filesystems only.

474902 / 46033776

102

Version 6.0.65.2


2 For a production file system mounted on VDM, if a new checkpoint was created with a name that has one or more spaces and with automount=yes, the checkpoint was created but remained unmounted. The following error was returned: Error 5005: Invalid argument failed to complete command.

The issue has been fixed by replacing every space with "_" character in mount path created from the checkpoint name.

476767 / 4665800

2 Disk Provisioning Wizard failed on NX4 system with the following error: "Not able to process templates. n Command failed, error code: 1, output: errorMessage:string="Timeout (180 seconds) waiting for state SS_TEMPLATES_GENERATED". "

This issue has been fixed by revising the code responsible for generation of hot spare templates.

483778 / 46920576

2 CLI command "nas_pool -list" failed with the following error: "Error 2237: Execution failed: Segmentation fault: Operating system signal. [VIRTUAL_DISK_TABLE.is_profile_supported] "

The issue has been fixed by initializing the storage_profile_ name to empty string when the name is not present in database record.

487182 / 47678224

3 File system auto extension failed when there was space in file system name.

This issue has been fixed quoting file system name in "rootnas_fs" command to allow spaces.

478780 / 46910078

103

Version 6.0.65.2


DART Core


2

Log event interface showed "Parsing error. Invalid XML” messages. The server_log showed following valid error messages, but with garbage characters at the end of each message. 2012-03-28 09:13:37: 150596354069: TIMESYNC: 3: ReadStream failed in time synchronization poll, error: _ŠÁ

. 2012-03-28 09:13:37: 150596354081: TIMESYNC: 3: Connection to the timeserver lost 2012-03-28 09:13:38: 73286942722: TIMESYNC: 1: 2: The NTP servers are not accessible or responding to the time service from [Wed Mar 28 09:13:37 2012].

The code has been corrected to display the accurate error messages. The log event no longer displays "Parsing error. Invalid XML” messages.

479218 / 46947888

1

The Data Mover panicked with the following message: DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG

The code has been fixed to avoid the panic condition. 483525 /

47383678

Deduplication


104

Version 6.0.65.2


1

Due to a race condition with DART's internal dedupe process, deduplication could be attempted on an inode that was deleted and recycled into another type of file. This could cause an internal file to be mistakenly deduped. The result could be loss of access due to blocked threads.

The race condition has been fixed. It is no longer possible for the DART deduplication process to accidentally compress the wrong inode.

487774 / 47818786

File System


1

When a file system was nearly full or reached a quota limit, a large copy from an SMB2 client could fail. If you retry the copy, it may complete successfully.

The Code has been fixed to correct the method to reserve and consume blocks during a copy from SMB2.

477292 / 46343472

1

Fsck panicked with header "PANIC in file: ../DeltaIO.cxx at line: 725 : DeltaIO_Read_CB::failed" due to bad block in the opaque data mapping pointers. This caused an extended DU situation.

The code has been enhanced to check for and skip bad blocks.

483028 / 46874118

1

When the amount of metadata changed, it caused the ufslog to become full. Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../staging.cxx at line: 439 : failed to write UFS log The result was that file systems should be checked before they could be mounted.

The code has been fixed to prevent the panic. When the Staging buffers are full, the current buffer will only be updated if the next buffer is unused.

490046 / 48019080

MPFS


1 A discrepancy may build between the du and df commands for an MPFS file system.

The issue has been corrected to free space

463971 / 45169514

105

Version 6.0.65.2


on MPFS mounted file system when files are deleted.

Network


1

DART panicked when the user issued a very large SNMP bulkcache request with the following message: DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG


478969 / 46919840

1

The following panic might occur while running Nesus security scanner: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x000020f310 Err code: 0 Target addr: 0x0000000028 ** Panicking Processor: 2 DART stack trace: 2860: 0xc2aef0 _ZN7DskDump9checkdumpEPKc+0x7a 28a0: 0xc2b0fe dump_return+0x4 28b0: 0x14d0b2 _ZN12Memory_Frame13setProtectionEm19Memory_Prot 28d0: 0x14d73d fault_dump+0x89 28f0: 0x14165d handle_fault+0x1d 69910: 0xd89489 _ZN13cPipeSimulate11checkPolicyEP13smb_threadC 69950: 0xde498b _ZN21cMsRPC_LsarpcSimulate15lsarCloseHandleEP1 699b0: 0xdf202e _ZN21msrpc_legacyOperationI21cMsRPC_LsarpcSimu

The code has been fixed to handle invalid MSRPC requests gracefully, returning an error.

484635 / 47497156

106

Version 6.0.65.2


NFS


1

It was possible to receive an illegal RPC reply with a NULL body and a status of "YP_TRUE" when using NIS with an NIS Server AIX 5.2 and/or AIX 5.3. The system expected the RPC message to contain at least 4 bytes and was not prepared to handle a NULL body RPC request, so it processed the request normally. This would result in storing data more than one byte into one byte space. That overwrite led to the following panic: DART X86_64 panic message: >>PANIC in file: ../malloc.cxx at line: 662 : Memory corrupted

The code has been fixed to check for a NULL RPC body to properly process the NULL message and reject the bad NIS response.

469561 / 45818200

NFSv4


1

The Data Mover panicked when a file system was unmounted while there was NFSV4 activity on it, with the following message: Page Fault Interrupt. Virt ADDRESS: 0x00006a4ba3 Err code: 0 Target addr: 0x0000000010 **

A spinlock has been introduced in the code to avoid reusing NFSv4 states which are being released.

476566

2

The premature revocation of LOCK states by the server caused BAD_STATEID errors. The server might revoke the LOCK state as soon as there was no file lock left attached to it; This did not comply with the RFC. The delay between the last LOCKU operation and the revocation of the LOCK state could vary between 0 and 5 minutes (nfsv4.cleaningInterval).

The Data Mover no longer releases the idle LOCK states associated with files which are open and idle.

484690 / 47472438

107

Version 6.0.65.2


Proactive Health Checks


2

If a CIFS server was present with WINS servers (used wins= option with server_cifs while adding CIFS servers), the server_checkup failed with the following error: Warning 17451974760: server_2 : VDM1: The WINS server '<IP address>' does not contain the correct IP address for the name '<CIFS server name>'. This one should be associated with <IP address>. The CIFS clients of this NetBIOS server might not be able to access the Data Mover.

The code has been fixed to enhance the IP address comparison function used during server_checkup.

479534 / 46324040

Quotas


2

While you had named streams in files, after running online quota check, quota block and inode usages would be less than the previous number.

The code has been fixed to calculate named stream and container files in quota check correctly.

466450 / 43457680

3 Querying user quotas on a tree produced no result, no error message from Linux client over NFS.

The code has been fixed to check the quota tree on/off status and return correct quotas.

473741 / 45791204

108

Version 6.0.65.2


3

Any query for tree quota usage following the modify tree quota operation returned 0. The same query would return values 24 hours later.

The code has been fixed; the query to return tree quota usage information following the modify tree quota operation now returns accurate results.

474357 / 46279178

1

There was no option to report capacities and usage based on file system and tree quotas that did not take user and/or group quotas into account.

A new option has been added to the param quota useQuotasInFsStat to enable reporting only file system and tree quota accounting to df queries from Unix clients. The existing values will behave as before. 0: (default) only the file system capacities will be reported to Unix clients 1: the file system and all quotas (user, group and tree) are taken into account when capacities are reported to Unix clients 2: file system and the tree quota are taken into account when capacities are reported to Unix clients

488023

1

When a large file was deleted, this was handled in the background as a deferred delete. If this process was interrupted by the file system being unmounted, it could result in tree quota accounting being incorrect.

The fix ensures the tree quota accounting is accurate.

493473

109

Version 6.0.65.2


Replication


1

Data mover panicked with the following message: DART panic/fault message: *** Invalid Opcode exception. Virt ADDRESS: 0x0000006bf6 ***


489235 / 47953422

Snapsure


2

Data Mover bootup or failover/failback could be slow due to checkpoint recovery when there were mergeDone checkpoints.These were caused by deleting or refreshing checkpoints which were not the oldest for the file system. The chunks for the deleted/refreshed checkpoints were merged into existing checkpoints and these chunks caused extra processing to be done during recovery after a reboot or failover/failback.

The checkpoint recovery code has been improved to skip reading the mergeDone chunk headers as it is not necessary.

408286

1

When latest ckpt was very large (several TB), it would take a long time (longer than 5 minutes) for reboot or failover.

The checkpoint recovery code has been improved, leading to improved reboot or failover time.

436839

1

Depending on the content in slot B, this would cause bad cg, fs corruption or btree corruption. The data mover panicked with the following message: DART panic/fault message: finishTake: bad magic in CG

The code has been fixed to avoid the panic.

472453 / 46122574

110

Version 6.0.65.2


1

Writing a PFS where a checkpoint was created caused a panic with the following message: "io not progressing". The recovery process was suspended after the panic. PFS on the replication destination side might be corrupted as a result of this.

The code has been fixed to correct the Snapsure recovery process.

483546 / 47403530

1 When replication started in reverse with HWM 0, it always returned an File_NoSpace error.

The code has been fixed so that only when the usage is greater than or equal to 75%, the replication reverse would fail, otherwise it succeeds.

491140 / 48120034

1 Repeated Data Mover reboots could be triggered during replication transfer or checkpoint merge on >2 TB file system. Any of the following messages could be displayed: Panic in file: ../DeltaIO.cxx at line: 802 : DeltaIO_Read_CB::failed Assertion failure: 'cv->unpackBtree32(curadd) >= startAddress/VOLBLK_SIZE' on line 3286 in file '../PagedBlockMap.cxx' >>PANIC in file: ../assert.c at line: 27 : Assertion failure: 'val != INVALIDNUM' on line 428 in file '../include/PagedBlockMap.hxx' >>PANIC in file: ../assert.c at line: 27 : assert failed

The code has been fixed to avoid panics by: • Avoiding the

insertion of invalid block numbers

• Adding checks to bypass invalid block numbers that may have been written previously.

497062 / 48857254

1 Rolling data mover panic occurred when a little amount of data wrote to PFS after a new CKPT creation and the corresponding save volume was corrupted. The following message was displayed:

The code has been fixed to avoid the panic condition by correcting the Snapsure recovery process.

497107 / 48872748

111

Version 6.0.65.2


DART panic/fault message: >>PANIC in file: ../PagedBlockMap.cxx at line: 2423 : Duplicate Mapping

Storage


1 Data mover panicked with "Volume has zero paths" . The following message was displayed:PANIC in file: ../include/volume.hxx at line: 414 : Volume has zero paths.

The code has been changed to not prune paths to lun on which scsi inquiry fails. This avoids the panic.

458436 / 44563854

1 A rare panic was encountered while processing a Tachyon fibre channel driver's Outbound Completion Message (OCM) with a set Invalid Entry (INV) error bit, although neither the OCM's SEST Programming Error (SPE) bit nor the VAL bit in the associated SEST were set. This condition violated the rules of the OCM's INV bit as specified in table 9-2 of the Tachyon FC Protocol Controller User Manual. The driver explicitly panicked the Data Mover (panic code = 0x01902009 == FC_EXM_INVALID_API_ERROR_STATUS with an additional error code = 0x00003002 == FC_CTLR_PROG_ERR) whenever an OCM's invalid entry bit was set without even looking at the OCM's SPE bit or the SEST entry's VAL bit.

The fix prevents this explicit panic within the Tachyon fibre channel driver (panic code = 0x01902009 == FC_EXM_INVALID_API_ERROR_STATUS) in the specific case (error code = 0x00003002 == FC_CTLR_PROG_ERR) where it violates the rules for the OCM's INV bit as specified in the Tachyon FC Protocol Controller User Manual.

459091 / 44655534

112

Version 6.0.65.2


1 DART panicked with a page fault. The following message was displayed: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000a69602 Err code: 0 Target addr: 0x0000000010 **

The offending code leading to page fault panic has been corrected.

470802 / 45934948

2 Deletion of the disk failed with error:- "Error 3501: Storage API code=24: SYMAPI_C_DEVICE_NOT_IN_DG. Could not select requested device(s) from the specified device group"

The issue has been fixed by changing the way the code determines the name of the device group the disk belongs to.

494402 / 48470488

SRDF

Severity

Symptom Description

Fix Summary

Tracking Number

1 RDF Set mode sync operation failed when Sync license was not enabled.

The fix will ignore the license error message to continue the restore process.

473453 / 45851944

1 nas_rdf -activate could be hung in nbs update operation. ps -ef shows "nd-clnt 6 7" like "/usr/local/bin/nd-clnt 6 7 server_2,server_3,server_4,server_5".

The fix should stop such a nd-clnt hung issue.

475565

System Management


1 When a security user was renamed, the home directory did not get renamed to reflect new name.

This issue has been fixed by renaming the home directory to match the new user name. If a directory with the new name exists, operation fails and an error is displayed.

90188

113

Version 6.0.65.2


2 No event notifications e-mails were received when an event notification was configured with Action 'mail' and Destination 'user' .

This issue has been fixed by correcting the software error that prevented email notification to be sent to the correct destination.

399760 / 37640588

2 After disabling a local user account, login into Unisphere through an LDAP authenticated account failed.

This issue has been fixed by correcting the user authentication code that handles disabled accounts.

459428

2 Incorrect alerts about hardware components appeared in Unisphere GUI after the actual faults were corrected.

This issue has been fixed by proactively deleting outdated temporary files used to determine statuses of hardware components.

466065 / 45410530

2 XML API failed to return all provisioned volumes.

The issue has been resolved by adjusting configuration parameter of Celerra Common Cache.

480447/ 45516028

2 Control Station periodically ran out of memory and rebooted.

This issue has been fixed by preventing Jserver from trying to access Symmetrix to check the threshold of the virtually provisioned pool.

486239 / 47586416

2 UIM could not discover the VNX and displayed the following error message: "HTTP error 405 method not allowed".

This issue has been fixed by moving the ECOM core files to /nas/var/dump/ partition, compressing the files and limiting the number of retained core files.

487238 / 47682672

2 Jserver returned disk volumes with Symmetrix serial number that had leading zeroes.

The issue has been fixed by stripping off leading zeroes from Symmetrix serial number.

488668 / 44892982

114

Version 6.0.65.2


VDM


2 The following error was seen in sys_log and the Unisphere event log: transfer_support_materials[26456]: The remote FTP server/directory cannot be reached. Verify your configuration in /nas/site/automaticcollection.cfg. Also, nas_checkup reported the following warning: Control Station: Check auto transfer status Warning HC_CS_18800050419: Cannot establish connection to FTP server sftphost sftp.emc.com Action : Command "/nas/tools/automaticcollection -verifyftp" failed to connect to FTP connection. Check the configuration in /nas/site/automaticcollection.cfg file. For more information, search the Knowledgebase on EMC Online Support as follows: 1. Log in to https://Support.emc.com and go to Support > Knowledgebase Search> Support Solutions Search. 2. Use ID emc221733 to search.

The issue was caused by a recent change in FTP server security, which caused the verification to fail. The code has been updated to work with the new FTP server security policy. See Primus article emc296211 for the temporary workaround for fixing the problem with the script until the system is upgraded to this VNX OE for File version.

491605 / 48169232

2 Checkpoint creation via XML API failed with error:- "The specified storage system '0' is not valid."

This issue has been fixed by modifying the XML API to correctly handle the configuration where both Control Station and the storage array have the same serial number.

493653 / 48326724

115

Version 6.0.60.2


2

The size of the security.evt log file increased beyond 512 KB when located on the root file system or on the root VDM when the GPO setting was set to a higher value, thereby filling up the root file system.

The code has been modified to limit the size of the security.evt file to 512 KB when located on the root file system or on the root file system of a VDM, regardless of the GPO settings.

473430 / 46236640

Version 6.0.60.2

Backup

Severity Symptom Description Fix Summary Tracking Number 1 VBB restore operation failed in

restoring directories with UTF8 characters.

The code was fixed to handle directories with UTF8 characters.

459567/ 43756396

1 The Data Mover panicked when the SavVol filled up during a greater than 60-day long backup and the oldest checkpoint was inactivated.

The code was fixed to protect against inactivated checkpoints panicking the Data Mover. The number of active versus inactive checkpoints was verified to be correct.

460751/ 44831934

1 The Data Mover panicked with the following message when DMA sent garbage data in env list during ndmp restore. PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG

The code was fixed to avoid the panic.

465634 / 45403186


Following are the fixed problems related to the Celerra Management Applications, that is, Celerra Monitor, Celerra Manager, WebUI, or Celerra MMC Snap-ins:


116

Version 6.0.60.2


2 Unable to join the CS to the storage domain. The CS appears to join properly and no errors are thrown but it never appears in the System List. This happened because systems on the same subnet with hostnames undifferentiated in the first 12 characters ended up with identical, non-unique System IDs.

The code was updated to generate unique System IDs, resulting in CS successfully adding to the storage domain.

435338 / 43225714

CIFS

Severity Symptom Description Fix Summary Tracking Number 1 Windows clients issuing a

DOS 'dir' command which included a complex mask resulted in all CIFS access to the file server being blocked. A forced panic of the Data Mover was needed to restore client access.

The code was changed to re-format DOS masks for optimal execution before they were submitted to the file server.

460364/ 44773644

2 When the CIFS server had many open files (likely more than 100000, depends on file names size), the open files enumeration over MSRPC (MMC, srvmgr) could not complete.

The code was fixed to prevent the cleanup task from running on in-progress active requests. This allowed the open files enumeration to complete.

462010/ 44848002

2 When connecting by CIFS to a qtree using the share of the root of the FS and the qtree path, the size of the qtree may be incorrect (the size of the FS is returned instead of the size of the qtree).

The code was fixed to return the correct size of qtree.

464557 / 45104472

2 Users could experience access issues to server's resource when domain was configured to use Resource Groups through Active Directory Federation Services (ADFS), as described in the Microsoft document: http://technet.microsoft.com/

The code was modified to handle credential resource groups correctly.

464981/ 45324904

http://technet.microsoft.com/en-us/library/cc737832(WS.10).aspx

117

Version 6.0.60.2


en-us/library/cc737832(WS.10).aspx

2 When a non server's administrator user attempted to connect C$ system share (access not granted), the access checking was not performed immediately, but access denied error was later reported on first request, when using the share.

The code was fixed to check for access immediately now.

466678 / 45510170

2 The server_stats and ”printstats cifs” did not report correctly for SMB2 TCP-connections and SMB1/SMB2 threads.

The SMB2 TCP-connections and SMB1 and SMB2 threads counters reported by server_stats were corrected.

468729/ No Clarify ID

2 The cifs acl.umask parameter was documented but not useable.

The code was enhanced to allow a CIFS user to create a file and get the Execute Unix mode bits set according the ACL of the file.

469523 / 45695180

2 Bad performance when copying several files from a Windows 7 client to a Data Mover share. This is because only one file was copied at a time.

The code was enhanced to improve performance.

473384 / 46207560

CIFS Virus Check


118

Version 6.0.60.2


Severity Symptom Description Fix Summary Tracking Number 2 When a file was being opened

for writing by one client, and another client was reading attributes (or extended attributes) of this file, an unnecessary virus checker scan was triggered for the file causing write performance degradation.

With the code change, the virus checker scan could not be triggered when the second client was reading attributes only on a file that was opened for writing from the first client.

463462/ 43901404

1 When an NFS user renamed a file on a file system with CEPP enabled for NFS, the Data Mover panicked with the following message: Page Fault Interrupt. Virt ADDRESS: 0x00007ec8f3 Err code: 0 Target addr: 0x0000000188

The code was fixed to avoid the panic condition.

463584/ 45067744

CORE

Severity Symptom Description Fix Summary Tracking Number 2 The Data Mover panicked with

the following message due to a memory leak when F-RDE was scanning a file system with a large number of hard linked files: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000159274 Err code: 2 Target addr: 0x0000000000 **

The code was fixed to avoid the panic condition.

459484/ 44220834

1 Under load, the Data Mover crashed with "Out of memory". The following message was displayed: DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory This could happen randomly

The code was fixed so that the memory used by the DBMS internal mechanism was pre-allocated at the initialization of the system to avoid the lack of memory situation.

460260/ 44774392

119

Version 6.0.60.2


Severity Symptom Description Fix Summary Tracking Number when the memory was fragmented.

Deduplication

Severity Symptom Description Fix Summary Tracking Number 1 Heavy CIFS access to

compressed file on an F-RDE enabled filesystem can cause excessive CPU load.

The code was fixed to avoid the situation causing excessive CPU load.

469525 / 45397734

DHSM


the following panic message when the user created a DHSM http connection and tried to modify this connection using an incorrect fs_dhsm command (-user option was used but no user name was provided). DART panic/fault message: *** GP exception. Virt ADDRESS: 0x0000aa783f. Err code: 0 ***

The DHSM command parser code was made more robust to avoid the panic.

422536/ 41631464

1 The user cannot read archived files when the stub file is restored from backup.

The code was enhanced in the DART kernel at the file system level to support the call that VBB restore operations was using.

440647 / 43018304

File System


120

Version 6.0.60.2


1 Data Mover server_2 panicked due to "Unnamed stream is wrong type". The release of DART internal resources for files containing (or once containing) alternate data streams is a multi-step release. DART incorrectly allowed an update to an inode that had already been freed. That inode had already been reused by another file and its contents were overwritten.

The code was fixed to avoid this condition during the release of files containing (or once having contained) alternate data streams.

440553 / 43857930

1 Using NFSv3 and lockd, could not unmount a file system on the Data Mover, resulting in a timeout.

Fixed a race condition which can occur when the lockd client uses several interfaces to set file locks on the Data Mover.

458487 / 44588032

1 When creating a new tree quota the Data Mover panicked with: PANIC in file: ../addrspac.cxx at line: 529 : Out of memory DART stack trace: e6e30: 0x1417d4 _ZN19Memory_AddressSpace10allocPagesEl12Memory e6e70: 0x242509 malloc+0x10b e6ed0: 0x243dcb _Znwm+0x1b e6ef0: 0x243df6 _Znam+0x4 e6f00: 0x812804 _ZN14UFS_FileSystem20treeQuotaOn_InternalEPcS0 e71c0: 0x815a26 _ZN14UFS_FileSystem16quotaOn_InternalEPcS0_+0x e7300: 0x786df6 _ZN14VFS_FileSystem7quotaOnE9quotaTypetPcS1_+0 e7350: 0x779f92

The memory allocation needed to enable a tree quota was split up so one large contiguous chunk is no longer needed. This will prevent the Data Mover panic when it has been up long enough for its memory to become fragmented.

459816 / 44708688

1 CPU usage intermittently spikes to 5% idle when there is high read activity on large read-only file systems.

The code was fixed so that the process to free inode block objects was

464911 / 45307774

121

Version 6.0.60.2


modified, so it consumes less CPU resources.

1 A corrupted NLM request caused Data Mover panic with the following message: Page Fault Interrupt. Virt ADDRESS: 0x000087d7fa Err code: 0 Target addr: 0 Datamover stack: #0 Lockd_Initial::start(this=0x360cdc40) lockdthr.cxx:1401 #1 Sthread_startThread_internal(ip=0x2d4464a60) sthread.cxx:698

Fixed the use of un-initialized pointer.

466579/ 45508014

2 When the ownership is changed for a file or directory containing Alternate Data Streams (ADS), the inodes for the ADS are not changed. This will cause user or group quotas to be incorrect.

Corrected the logic error to make the ownership of ADS consistent with the file/directory.

467158 / No Clarify ID

1 The Data Mover hung in a rare situation while there were 3 tasks running at the same time: quota check or quota on task, IO task, and snapsure task.

The code was fixed to avoid the hang.

468422 / 45715556

1 Threads become blocked eventually leading to a loss of access. This can happen when a file or directory containing a compressed named stream is deleted. In DART 5.6.47 and above, DART 6.0, and VNX OE for File 7.0, compression of named streams is not performed so this issue is only possible on a file system where dedupe was enabled on 5.6.46 or earlier.

The code was fixed to change the locking that is required when deleting a named stream to avoid the potential of a deadlock.

469851 / 45822506

122

Version 6.0.60.2


1 Data Mover panics with: PANIC in file: ../ufslog.cxx at line: 2728 : filesystem log transaction record overflow This happens due to a rare case when a file system operation results in a ufslog transaction larger than the maximum allowed.

The code was changed so that the number of blocks which can be included in a ufslog transaction have been reduced to ensure that the total transaction size will not exceed the maximum allowed.

476163 / 46597972

FLR


1 Data Mover panicked with the following message: DART panic/fault message: >>PANIC in file: ../rawirp.cxx at line: 607 : Write Verify failed for FLR data


461016/ 44839716

FSCK

Severity Symptom Description Fix Summary Tracking Number 1 If a PBR check was required

during FSCK or if quota recalculation was required when mounting the file system, the inode scan routine did not print any progress message making the user believe that the operation was hung.

The code was fixed so that the inode scan could print the progress messages.

463902/ 45121204

Network

Severity Symptom Description Fix Summary Tracking Number 1 If IPv6 was configured in the

network environment, but IPv6 interface was not configured on the Data Mover, the Data Mover could periodically panic

The panic was triggered by invalid sanity checking code, which did not expect a destination

440268 / 43805094

123

Version 6.0.60.2


Severity Symptom Description Fix Summary Tracking Number with a panic dump header similar to: *** Page Fault Interrupt. Virt ADDRESS: 0x0000489c28 Err code: 2 Target addr: 0x0000000010 **

address to be IPv6, and source address to be IPv4. This sanity checking code has been fixed.

1 Data Mover panicked with the following: PANIC in file: ../bufpool.cxx at line: 425 : couldn't get a free page

If a customer joins a CIFS Server to a Windows domain that is using a non-Windows DNS Server, the subsequent secure (Kerberos) DNS update from the Data Mover will fail because the DNS Server is not registered in Kerberos for the domain. If the Kerberos credentials for the CIFS Server are cached in a certain order, memory leaks will occur due to a credential being retrieved from the cache and selected as a 'match', then not used.

Additional calls to free the credential contents were added to prevent the memory leaks.

460236 / 44786942

1 The Data Mover panicked due to experiencing IP fragmentation and any of the fragments are zero length. The following message was displayed: DART panic/fault message: Page Fault Interrupt. Virt ADDRESS: 0x00003d49a8 Err code: 2 Target addr: 0x0000000000

The code was enhanced to avoid UDP and optimize infrastructure to avoid IP fragmentation.

470446 / 45912724

124

Version 6.0.60.2


NFS

Severity

Symptom Description Fix Summary Tracking Number

1 The Data Mover panicked with the following message: Page Fault Interrupt. Virt ADDRESS: 6638df Err code: 2 Target addr: 4 There were server logs like "NFS: 4: xid XXXX in progress - reset after drop more than 8 times "


470457 / 45931264

NFSv4


the following message: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x00006a7fc0 Err code: 0 Target addr: 0x0000000000 **


460992/ 44859406

2 The AIX client hung when using NFSv4.

The Server could no longer grant a read delegation if the open access mode was read + write. If a write delegation could not be granted, no delegation would be granted at all.

464549/ 44422786

1 All NFS threads were blocked causing NFS access to be lost. This could have happened due to an edge condition with NFSv4 when verifying the attributes for an NFSv4 operation leads to an infinite loop.

The code was changed, so that the conditions which could trigger this infinite loop were fixed.

465354/ 45380486

1 The NFSV4 server returned the error NFS4_RESOURCE when

The code was fixed to release an open

465606/ 45378140

125

Version 6.0.60.2


Severity Symptom Description Fix Summary Tracking Number there were too many open owners. This could also happen if there were few opened files. The reason was that open owners were released only after a lease period to allow close replay detection.

owner when it had no more files opened. This avoided the issue.

Platforms

Severity Symptom Description Fix Summary Tracking Number 3 HTTP 413 “Request Entity Too

Large” page might return unsanitized client header.

Changed webserver configuration file to use a static error message.

355867 / 33486008

2 The connectemc daemon hung and stopped sending the callhome files.

A solution was implemented to detect and restart the connectemc main process if it was hung, and added a regular restart mechanism of the connectemc main process as the last recovery resort.

453530 / 44047694

2 By default, the web server on the Control Station allowed weak ciphers and SSLV2 to be used.

Changed the web server configuration file to disallow weak ciphers and SSLV2 by default

456362 / No Clarify ID

2 syslog messages report blank for Username fields in case of local user when AUDIT_Tool is run.

New code was added to read local user if no domain user found, and display the local user in AUDIT_Tool output.

470581 / 43023590

Quotas

Severity Symptom Description Fix Summary Tracking Number 2 ”User Name” column on quota

information page in Unisphere showed the User ID and not the Domain\Username.

The code was fixed to display the user information correctly.


126

Version 6.0.60.2


SnapSure


1 The Data Mover panicked on ckpt fs with fs corruption; or on snapsure layer with Duplicate Mapping. The following message was displayed: DART panic/fault message: >>PANIC in file: ../DeltaVol.cxx at line: 3844 : doSavVolIOVec

Code was fixed to avoid the panic situation.

439471 / 43674774

Storage

Severity Symptom Description Fix Summary Tracking Number 1 "server_mt -f [device] status"

fails when connected to DataDomain VTL devices.

Code was fixed so that server_mt returns the correct status.

392665 / 36965502

2 The NDMP Backup failed with the following medium write error message on direct attached LTO5 drive with append only feature : ---------- server log--- CAM: 3: cdb: 0a 00 00 04 00 00 00 00 00 00 00 00 | SCSI ERR ON #3200 camstat 0x84 bstat InvalidOperation | scsi stat.skey.ascq 0x02.07.5a02 Check condition | skey [ ] Write protect |____________________

The code was changed to add support for 16 byte SCSI commands for tape/robot devices.

431941/ 41648218

127

Version 6.0.60.2


Severity Symptom Description Fix Summary Tracking Number ascq Operator selected write protect

System Management

Severity Symptom Description Fix Summary Tracking Number 2 Abort of "server_ping -send"

command with Ctrl+C hangs serial connection to Control Station.

Code was modified to protect output to console operation from being interrupted by external signal, and the interrupt signal is delivered after the operation completes.

406307 / 39377292

2 File Events filtering by severity does not work correctly on Unisphere Event Logs page: when user requests "Info" events only, both "Info" and "Warning" server events get displayed. And when the user selects "Warning" events only, no server events get displayed at all.

The code was corrected to ensure that server events get retrieved according to the user specified severity level.

439164 / 42989446

1 Upgrade failed with error message "Error IM_14505279794 : Failed to setup the following local storage devices ..." due to ongoing RAID group defragmentation on storage array.

The code was changed to remove obsolete 'exdisks' check performed while running 'nas_storage -check -all' command.

458484/ 44421782

3 Server_export command failed with error message ”Invalid argument” when a host FQDN contained "sec" clause.

The code that processed export options was modified to ensure that "sec" clause in FQDN did not get process as "sec=" export option.

460366/ 44678140

128

Version 6.0.60.2



2 Errors like the following appeared in the server log of a standby server: 2012-01-09 02:35:26: 13160087552: SVTL: 3: Command failed: poolsInfo 2012-01-09 02:35:26: 26044989440: SVTL: 6: SvtlServer() --> svtl_rename(/svtl, /.etc/svtl) failed: Stale handle 2012-01-09 02:35:26: 13160087552: SVTL: 3: mkDir(/.etc/svtl/locks/libraries) failed: IO_Error 2012-01-09 02:35:26: 13160087552: SVTL: 3: SvtlServer() --> pTlusMutex->Init(/.etc/svtl/locks/libraries) SVTL_STATUS_ERROR 2012-01-09 02:35:26: 13160087552: SVTL: 3: SvtlServer::Alloc(/.etc/svtl): SVTL_STATUS_ERROR

The code was modified to disallow sending VTLU queries from Control Station to standby server.

465968/ 45378244

2 Control Station exhibits high CPU load.

The code was fixed to reduce Control Station CPU load by limiting periodic filesystems queries initiated from Jserver

470280 / 45558122

UFS

Severity Symptom Description Fix Summary Tracking Number 1 Failed to mount the file system

with the following message: ”Mount not allowed: too many mounts on this server”.

The code was modified to handle this situation.

165703/ 33316276

129

Version 6.0.55.3


Severity Symptom Description Fix Summary Tracking Number 1 CPU usage spikes of 100

percent lasting for several seconds occurred when clients accessed tens of thousands of directories.

The code was changed to reclaim memory used in directory processing more efficiently.

457456/ 44256920

1 The Data Mover panicked with panic header ”Assertion failure: 'aclHdr.getHdr()->revision == ACLRev” if ACL database was full (1 million different ACLs) and there were concurrent threads trying to allocate a new ACL.

The code was fixed so that the solution eliminated the condition where two threads attempt to allocate the last ACL entry.

467142/ 45564250

Version 6.0.55.3


Following are the fixed problems related to the Celerra Management Applications, that is, Celerra Monitor, Celerra Manager, WebUI, or Celerra MMC Snap-ins:


2 Celerra Monitor failed to start with error: "Component <serial_number> is not accessible to this user"

Data types in Jserver have been modified to fit Symmetrix director configuration attributes without truncation

420801/ 41209492

2 Unisphere and Celerra Monitor did not display filesystem capacity and inodes utilization information for a filesystem name that included one or more space characters.

The code has been modified to correctly handle filesystem names that included space characters.

432871/ 42832926

2 A Data Mover failing over generated a very large number of problem reports which filled up the Jserver partition. This occurred due to a space issue under /nas/jserver partition.

Jserver code was modified to stop producing problem reports when stale data was received from Common Cache. Jserver now records the record problem in the log file.

439763/ 11111111

130

Version 6.0.55.3


2 Numerous warning messages similar to the following in sys_log files: CS_PLATFORM:ADMIN:WARNING:18:::::dskMon[1245]: FS /dev/mapper/emc_vg_lun_5-emc_lv_nas_var_emcsupport mounted on /nbsnas/var/emcsupport filling up (at 100%, max = 90%).

The code has been fixed to prevent the emcsupport partition from filling up. The issue that caused a background process to run infinitely has been corrected.

456533

1 Events were missing or the entire view was empty when viewing events from the Event Log Refreshing the view restored some or all of the events. More than one refresh caused events to be dropped or the view to empty.

The code has been fixed to display events correctly.

462851

1 On very large (greater than 10 TB) MGFS filesystem after a large amount of data is migrated, the system started appearing unresponsive.

The code has been modified to initialize cg bitmap for MGFS filesystems to avoid unresponsive behavior.

455240 / 44242956

CIFS

The following entries describe fixed Common Internet File System problems:


2 In Windows Vista®, a new SID, OwnerRights, can be used in ACEs to control the rights on securable objects. The File OE did not support the new Vista OwnerRights SID.

The code has been changed to support Vista SID. 368438 /

34831146

1 Data Mover panicked when Usermapper service received unexpected packet. Typically when security scanner sent invalid RPC to the usermapper service.

The code has been hardened to validate and reject incoming requests that are not formatted in the correct manner.

404497/ 39107628

2 An open file could not be forced closed on VNXe based storage systems.

The code has been modified to allow open files to be forced closed upon request.

433118

1 All the storage system file systems were visible and accessible through the C$ share.

The code has been modified to add a parameter (cifs.admin.shareC_NotCreated)to hide C$. The default for the parameter leaves C$ visible.

434264

131

Version 6.0.55.3



Disabling C$ visibility will not allow users to browse the different file systems from Microsoft MMC when a share is created.

2 With Windows 7 and SMB2 client access, if a Word document had its archive bit un-set in DOS attributes, a modification in the Word document did not reset the archive bit.

The code has been modified. Archive bit now works as expected with Word in Windows 7 and SMB2.

434306/ 43096796

1 The Data Mover panicked when an application made a very large number of accesses to the same file from many different Windows Vista and Windows 7 SMB2 clients that used SMB2 file lock request operations. The message displayed is “Run out of free blocks in Buffer_CachedPool!"

The code has been modified to correct the memory leak in the SMB2 file locking mechanism.

436698/ 43406710

2 Users experienced a WERR_INVALID_PARAMETER error when attempting to configure shares from SQL clustered 2008 R2 server. This is because Celerra was rejecting the NetShareSetInfo request if the shi502_netname field was not populated.

If a netname is missing in a NetrShareSetInfo request, the code uses the share name from qualifier to prevent this error.

436846/ 42935986

2 The command server_checkup generated an error in its report when executed on the FTPD test with a standard default configuration of FTPD. When the default persona of the FTP server is empty (no certificate and no private keys), then it generates the error “no available certificate for persona 'default'”. This error message is not correct.

The code has been modified. When the command is executed the validity of certificates in the persona is checked only if SSL is allowed, required, or enabled (that is,not disabled).

440016/ 43715054

2 In quota reports, a discrepancy was observed in the reports shown in CLI and GUI quota screen for a approximately 234 users. This was in context with domain migration.

The code has been fixed to avoid missing entries in the quotas report.

440448/ 43731696

132

Version 6.0.55.3



1 The Data Mover panicked when the Codenomicon/Defensics security scanner sent invalid SMB2 packets. The following message was displayed “>>PANIC in file: include/stream.hxx at line: 1130: datab:freeRef: db_ref <= 0”

The code was fixed to protect against corrupted SMB2 commands. The code now rejects invalid commands and returns an error to the client.

453320/ 44042010

2

"The permissions are incorrectly ordered“ error message was displayed when the CIFS users opened the security tab for a file or folder. This was due to improper sorting of deny and access permissions by NFSv4 ACL.

The code was changed so that NFSv4 ACL correctly sorts deny and access permissions.

458223/ 44455658

2 Windows 8 clients were unable to map EMC server shares. The net use command failed with system error 2148073478.

The code has been fixed to support Windows 8 compatibility. This ensures that the Data Mover selectively checks an SMB2 request and signs the response even if the server was not configured to force signing.

459810

2 A Windows user trying to create a new file (right click in Window Explorer) did not get a new option if the share they were working on was mounted on a network drive;file system is a writeable snap and client OS is Win7/W2K8

The code has been fixed. When access to share is checked, we return Read-Only if the file system is a snap and not a write-able snap.

461052

2 Copying files from MacOS X Lion to a Celerra CIFS share failed when the UNIX type SID had the form S-1-15-88.

The code was modified to accept ACEs with SID 1-5-88 without returning an error.

437759 / 43336414

2 When using the Widelinks feature all shares for the CIFS server will have the DFS root attribute set, even those without any Widelinks under the share. This was found to cause an issue with Windows7 clients using offline caching mode because the DFS attribute affects which referral it selects.

A new NT registry key HKEY_LOCAL_MACHINE/Software/EMC/Widelink/Global (Set per VDM) has been added as a choice if Widelinks is globally supported or per share using a new attribute. By default, HKEY_LOCAL_MACHINE/Software/EMC/Widelink/Global will be set to 1 which will result in keeping the existing

438207 / 43020618

133

Version 6.0.55.3



behavior. When this key is set to 0 the Widelink attribute will need to be specially set on all shares containing Widelinks. The new Widelink attribute will be supported on share type (like for ABE and Offline cache). This attribute will have to be set using the type option on the server_export command: server_export -name myShare -option type=Widelink /myPath

CORE

The following entries describe fixed problems related to Celerra Data Movers:


1 The Data Mover panicked when a file system mount point included special character "=". The root cause of the Dart panic is NULL fs volume ID (fsVolId). It returns from parsing Mount Command string with more than one pre-defined separator "=".

The code has been modified to handle special character “=”.

431762 456060/ 42742496

1 The Data Mover panicked when a path larger than 256 characters is entered in a command.

The code was modified to handle paths greater than 256 characters in length.

436681/ 43287434

134

Version 6.0.55.3



2 Creating a checkpoint on the Control Station while the production file system was mounted on a Unix/Linux client did not allow that client to immediately cd into the .ckpt directory from the root of the file system, and/or all the checkpoints were not visible from the .ckpt directory at the root of the file system. Once the file system was unmounted and mounted, you could then cd to .ckpt, and all the checkpoints were visible from the root of the file system.

The code was fixed so that the checkpoints were assigned correct values when they were created, and all checkpoints and the .ckpt directory were visible to the client from the root of the PFS immediately after creation.


Deduplication

The following entries describe fixed issues related to Deduplication:


1 The CIFS and NFS server became unresponsive and forced a reboot when using deduplication.

The deadlock was fixed by changing statistics logging. The logging caused a lock to be acquired in the wrong order.

440939/ 43864948

1 The Data Mover panicked when Deduplication was scanning a file system containing a large number of hard linked files. The DART panic/fault message was “Page Fault Interrupt. Virt ADDRESS: 0x0000159274 Err code: 2 Target addr: 0x0000000000”

The memory leak was fixed. The code now releases allocated memory after the scan.

455471/ 44220834

1 Following a datamover panic and failover, users were unable to open deduplicated files. If a problem was found with the internal deduplication policy engine the state of deduplication would be set to Off. Under normal circumstances when deduplication is set off all files are recalled, however in this situation the state was set off but no recall was performed. This resulted in users being unable to access any deduplicated files.

Deduplication will be suspended which means no new files will be deduplicated while still allowing access to any previously deduplicated files.

457093 / 43892970

135

Version 6.0.55.3


File Systems

The following entries describe fixed File System problems:


1

The Data Mover panicked while creating a Read/Write check point filesystem. The message shown was: “PANIC:>PANIC in file:../addrspac.cxxat line: 529:”

The code was modified to correct a memory leak in file system initialization. 436316

1 A Data Mover could deadlock if FS checkpoint or any other tasks that needed FS pause while iSCSI LUN/snap deletes were going on in the background.

Locks were acquired in the proper order to avoid the deadlock situation.

459827/ 44694588

iSCSI

The following entries describe fixed issues related to iSCSI LUNs:


1 iSCSI LUNs that resided on the same filesystem were erroneously reported to the iSCSI initiator with the same WWN when the file system was un-mounted on boot-up. If the file system issue was corrected and the file system was re-mounted, the iSCSI LUNs were still reported to the initiator with the same WWN until the Data Mover was rebooted.

The code has been fixed so that the iSCSI service will not register a WWN for iSCSI LUNs whose file system is un-mounted and will not report those LUNs to the iSCSI initiator. Once the file system issue is corrected and the filesystem is re-mounted, the iSCSI service must be restarted by running the command "server_iscsi server_x -service -stop" and then the command "server_iscsi server_x -service -start" -or- by rebooting the Data Mover.After the iSCSI service is restarted, it will compute a proper WWN for the affected iSCSI LUNs and report them to the iSCSI initiator.

453767/ 44110140

Installation, Configuration, and Upgrades

The following entries describe fixed issues related to installations, configurations and upgrades:

136

Version 6.0.55.3



1 Earlier releases used ConnectEMC 2.6 software to deliver callhome messages to EMC. The outgoing mail functionality of this version issued duplicate HELO/EHLO commands. This was in accordance with RFC 2821, however, some mail servers do not support this (especially older mail servers).

This release uses ConnectEMC 3.0.1, which addresses the callhome issue.

377477/ 378198

1 Biweekly health checks ran for a very long time and consumed a lot of CPU cycles if "server_checkup" reported more than 40 errors or warnings for one Data Mover.

The health check code has been modified to avoid this situation.

461161 / 44639742

NFS

The following issues describe fixed issues related to the Network File Server:


1 The Data Mover panicked when the NFSv4 client was revoked and requests to confirm its callback path had not been cancelled. The message displayed was “path is unavailable(frozen) or invalid”

The code was modified to eliminate the race condition by making a check to verify the callback path has been cancelled before the NFSv4 revocation takes effect.

434416

1 The signature verification on a secure NFS request to delete the session security context was incorrectly processed, resulting in an auth_error response to the client. In the case of a Mac OS X client, the error caused the client to go into a rapid loop of creating and deleting additional sessions, which severely affected the Data Mover performance.

The signature verification processing code was fixed to return the correct response.

453465/ 43760246

1 When using NFSv4, an unexpected error status was returned by the server (BAD_STATE_ID), making the file unaccessible.

The code has been fixed to return the correct error status (OLD_STATE_ID).

456022/ 44309440

1 After running chmod for files with NFSv4 ACL setting, inherited ACL settings got lost. INHERITED flag was dropped by NFSV4 ACL manipulations, because it was not returned to the client.

The code has been fixed. This flag will be correctly translated in NFSV4.

453335 / 43875390

137

Version 6.0.55.3


1 Loss of NFS access after changing special mode bits (sticky, suid etc ...) from NFSv4. This issue is only possible when Windows security auditing is enabled and the file system is mounted using Mixed access policy. The server log will show messages like: KERNEL: 3: 3: ThreadsServicesSupervisor: Service:NFSD Pool:NFSD_Exec BLOCKED for 361 seconds: Server operations may be impacted.

The code has been modified to eliminate the potential deadlock when changing special mode bits (sticky, suid, and so on) from NFSv4.

455710 / 44297386

Replication

The following issues describe fixed issues in this release related to Celerra Replicator, SnapSure, and Checkpoint:


1 When the Data Mover panicked during a replication, the create BLO reported success to the user without writing the session signature to the sav volume. This later caused a problem for stop/start BLO, when a session was lost.

The code has been modified to attempt to re-write the session signature after a panic interrupts a replication. If the re-write fails, the BLO will be rolled back and a CCMD message will be issued to the user.

426987

2 The create remote replication command failed when a session on the source had the same FSID as a session on the destination.

The code has been modified to reject the creation of a new session between a source and destination that is not unique.

431722/ 42744966

1 When a checkpoint was restored, it created a temporary checkpoint and snapped restored data. If there was not enough space, the restore failed leaving the filesystem inconsistent and replication inactive. The replication stop operation would attempt to restore a checkpoint to bring the destination to a consistent state. The restore operation failed as the destination savvol was full and could not be extended.

The issue has been fixed. When replication aborts and tries to restore the PFS, the special restore , which will not create temporary checkpoint and not allocate any other chunk, is launched. 398615 /

38346418

138

Version 6.0.55.3


1 A queued ReplicationV2 session checkpoint operation (create/delete/refresh/restore) on the destination experienced long delays when the SavVol needed to be extended in order for the transfer to complete.

The code has been changed to allow extends to continue, And, to reject any queued checkpoint operation requests. 416453 /

40756596

1 DART panicked with the following panic header: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 820db2 Err code: 0 Target addr: 0 **

The code has been fixed to avoid the DART panic.

421194 / 41485422

1 DART panicked with the following panic header: DART panic/fault message

The code has been fixed to avoid the DART panic.

439878 / 43748484

UFS

The following entries describe fixed UNIX File System issues:


1 The Data Mover panicked due to an out of memory condition under a high work load and with many active file systems. The condition was caused by Cylinder Group (CG) buffers being held in-use and remaining indefinitely memory resident. The following message was displayed “>>PANIC in file: ../addrspac.cxx at line: 289 : Out of memory”

The CG caching algorithm code was changed to more efficiently handle the aging of CG buffers.

394853/ 37939478

1 The Data Mover panicked with the following message: DART panic/fault message: >>PANIC in file: ../ufs.hxx at line: 1377 : UFS hashRemove: not on hash chain

The code has been modified to avoid the panic

421397 / 41500892

1 Data Mover panicked with DART panic/fault message: Assertion failure: 'this->back == this' on line 551 in file ”../vfsnode.cxx” >>PANIC in file: ../assert.c at line: 27 : assert failed

Fixed the race condition on the mount point when file system was being unmounted and another thread is still referring the mount point.

434499 / 43149914

139

Version 6.0.55.3


1 Watchdog panicked while enumerating file-system stats using server_stats command.

The code has been modified to avoid the watchdog panic.

453141 / 43990760

1 When the system is short of memory, a rwlock may not be released in the error path if the I/O is in the uncached IO path. This causes NFS threads to hang.

The code has been fixed to release this lock in the error path.

461028 / 44875620

1 The Data Mover panicked and reported that a file system requires an File System Consistency Check (FSCK). The panic may report a corrupted directory entry or an invalid indirect block.

The code was changed to correct invalid free block recoveries at the time of Data Mover recovery, avoiding further Data Mover panics.

466190 / 45446328

Control Station

The following entries describe fixed problems related to the Celerra Control Station:


1 Slow overall system performance was observed with nas_pool -query being issued from JServer. The poor performance could be seen in several different ways including RDF continuously failing over, slow GUI response times, certain task starvations, etc

The code now provides a number of performance optimizations to improve performance.

422395 / 41595948

2 “dbchk –p” command failed when executing "symmir list", and gave out error “Error: File system with id /nas/symcli/bin/symmir list is found in .server_config mountdisplay's output but does not exits in /nas/server/{1}/mount file.” The error doesn't describe the actual problem.

The error message was replaced with ”The exit status is neither “0” nor “1” while running command symmir list.”

454984 / 44202128

140

Version 6.0.55.3


Network

The following entries describe fixed Network-related problems:


1 BCM 5715 1GigE NIC ports intermittently showed link bounce.

Changed the default behavior of the BCM NIC recovery mechanism to not reset the 5715.

432345 / 42831902

System Management

The following entries describe fixed issues related to system management:


2 When control station is in BRST timezone, if schedule fire-off time falls into the transition time of BRST to BRT, scheduler operations get hung.

The code has been fixed. Scheduler no longer hangs.

436472 / 43375724

Quotas

The following entries describe fixed issues related to quotas:


1 When a CIFS client issues quota on while a online quota check task is running, the Data Mover panicked with: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x000083f4bc Err code: 2 Target addr: 0x0000000000 **

While logging a server log message indicating that an online check was already in progress, the code was dereferencing a null pointer

The condition is now handled correctly.

453514 / 44008566

141

Version 6.0.51.6


Version 6.0.51.6


The following are fixed problems related to the Celerra Management Applications, that is, Celerra Monitor, Celerra Manager, WebUI, or Celerra MMC Snap-ins:


2 Following a Data Mover failover, VDM CIFS servers were not listed in Unisphere. They were visible through the CLI.

CIFS servers created on a VDM are now visible in Unisphere after a Data Mover failover/failback.

406031 / 39311966

3 Persistent failures of APL tasks over time, due to faulty logic validating UPS configuration (expecting 2 UPS configuration when only one exists) resulted in accumulation of large amount of tasks related files in /nas/tasks directory. Eventually 100% of file descriptors were exhausted. NBSnas becomes non writable, and nas services will fail to start.

The logic has been corrected and these APL task failures will no longer occur.

411571 / 40109354

2 In Unisphere, when managing the Usermapper configuration and uploading "User Mapping File" or "Group Mapping File", or both with the "replace" option caused one of the files contents to be lost. As the result Windows users may lose access to their files.

Modified Mapping Files processing algorithm to ensure that using the "replace" option does not cause loss of the contents.

412850

2 Automatic DHSM connection creation and connection verification for Rainfinity FMA using XML API failing due to case sensitivity of share queries. Users could not query the location if they didn't specify the share names in the actual case.

The share name comparison have been changed to case insensitive.

422160

142

Version 6.0.51.6



1 When there was a large number of objects in user and group quota pages, user experienced delays in screen painting. The screen refreshes itself several times, and each refresh causes significant delay which appears like a hang to the user.

Fix involves some design changes in the way tables are built and updated.

396840 / 37832254

2 User and group names were not displayed on the user and group quota page. Instead user and group IDs were displayed.

Changed the APL to support the correct schema reporting of the SIDs value for user and group quotas.

397447

2 Jserver reported incorrect port WWN for SP ports when a setup has more than 8 FC FE ports per backend.

The fix is to hash ports info based upon the entire port WWN instead of just the "port number".

420482 / 40714894

2 Creation of a new notification in Unisphere fails when e-mail address for notification contains '&' character. The error messages says: "Celerra Monitoring Agent could not be contacted".

The code now properly handles special characters, like '&', '<', '>', etc. when included in the e-mail address for notification.

432113 / 42573498

2 When using NTP QFS software with Celerra, Celerra Manager login page failed to load. That problem disappeared after a Control Station reboot, however returns back after several days. This was due to QFS software being disconnected abruptly and leaving HTTP socket open.

Improved code that handles abruptly disconnected XML API client. Introduced a limit on the number of outstanding XML API requests.

431655 / 42590566

2 Unisphere showed only one event starting 31st October 2011. The alert_log file was checked, it was normal and contained multiple entries.

The APL filtering error that was causing this issue has been corrected.

454109 / 44118074

1 The Schedule Name column was not visible in the view even if the SnapSure license was enabled in Unisphere.

Changed the string concatenation operations to ensure that when "snapsure" license is enabled, user

454360

143

Version 6.0.51.6



can see the Schedules column.

CIFS



1 If a share path was being used by another share command while the path was being relocated, the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 359: Memory corrupted.. The stack trace showed: 0xe94ff740: 0x141278 waitForReboot+0x90

0xe94ff760: 0x141501 fault_dump+0x67

0xe94ff780: 0x1413e9 PANIC+0x29

0xe94ff790: 0x200731 free+0x6f

0xe94ff7c0: 0x200e34 _ZdlPv+0x10

0xe94ff7d0: 0x8aecce

_ZN12volatilePathIcED0Ev+0x2e

0xe94ff7e0: 0xc89ff2

_ZN14abstractStringD2Ev+0x20

0xe94ff7f0: 0xc8a13d

_ZN10pathStringD0Ev+0x15

Code was modified to ensure that the share path is not being used or modified when the path is relocated.

431284 /42686198

144

Version 6.0.51.6



1 When the virus checker service was restarted with new masks and exclusion rules, these changes were not applied to files that were open. This resulted in the virus checker scanning files with an excluded extension.

Code now checks whether the virus checker configuration has changed before performing a scan of a file.

435234 /43216602

1 When there were several open resources on a Domain Controller in trusted domain which no longer responded, each SMB thread entered a 20-second timeout, potentially creating a CIFS outage. During CIFS access outage, the server_log was logging "DOMAIN_CONTROLLER_NOT_FOUND" events. In addition to this, the Data Mover was sending MSRPC Close requests when the TCP connection was already dropped.

The code has been fixed and Data Mover will no longer cause CIFS outage under such conditions. The Data Mover will no longer send unnecessary close requests when connection is already dropped.

372385 /35027576

1 When a Data Mover requested user’s group memberships, a Domain Controller, for some unknown reason, had not sent the primary group of a user. This resulted in access being denied to files that rely on group membership as the group membership was not stored in internal credential.

The fix adds missing primary group to internal credential to prevent this condition.

407896 / 39421448

2 File system outage occurred when enumerating CIFS quotas with slow mapping mechanisms. User lost access to his shares while the quota entries window was populating CIFS quotas window.

Code was modified to reduce the file system locking time when enumerating users' quotas from CIFS.

412335 /39230656

1 When certain kinds of malformed CIFS/SMB or NFS packets were sent to a Data Mover’s CIFS/NFS server or that of a Celerra, the Data Mover panicked, resulting in a denial of service for legitimate users. One example of this was seen when Codenomicon's Defensics v3.12 (www.codenomicon.com)

Improved handling of invalid data being sent in fields within an NFS or an SMB2 request.

414090

145

Version 6.0.51.6



vulnerability scanner was run against a Celerra Data Mover.

1 With widelinks enabled and many CIFS accesses to a share from one client, the Celerra CIFS server incorrectly returned STATUS_OBJECT_PATH_NOT_FOUND for a request for a valid path. This was due to a race condition where multiple threads were checking a variable to determine whether it was a widelinks connection.

The check for whether it is a widelinks connection is now done only once during tree connect or mount of a file system and not on each protocol request which was leading to the race condition.

414865 / 40421472

146

Version 6.0.51.6



1 The Registry.pol file is used to maintain certain GPO settings on a Windows domain controller. User had a (possibly corrupted) Registry.pol file where fields (in a single entry) were all zeroes, so the pointer to the data field was set to NULL. When the NULL data pointer was subsequently dereferenced, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000f76e6a Err code: 0 Target addr:

0x0000000001 **. The stack trace showed: 0760: 0xc1ad80 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xc1af8e dump_return+0x4 07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 07d0: 0x14ce5d fault_dump+0x89 07f0: 0x14162d handle_fault+0x1d 60d1b0: 0xf77f2d _ZN6SysVol19getRegistrySettingsERK6Strin gRP8L 60d320: 0xf7beb0 _ZN6SysVol14getGpoSettingsERK6StringRP12 gpo_s 60d770: 0xf52346 _ZN9DomainGpo18updateVdmGpoServerERK6Str ingi+=

The data length field value is now checked (to be greater than 0) before dereferencing the data pointer.

422696 / 41650866

2 One of the Group Policy settings used by the Data Mover is the Maximum machine account password age. When this setting is defined or changed in the domain,

The code was changed so that the default setting (0) is applied to the applicable CIFS Servers, when the

418168

147

Version 6.0.51.6



it is updated and used by each CIFS server. If the setting was set to not defined, the applicable CIFS servers were not updated with the new setting value until CIFS was restarted.

setting is changed to no longer defined in the domain.

1 When an administrator tried to list open files the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: bc9a8e Err code: 0 Target addr: 24 **. The stack trace showed: 0x2c3b91c: 0x14124c waitForReboot+0x90 0x2c3b93c: 0x1414d5 fault_dump+0x67 0x2c3b95c: 0x13a398 handle_fatal_fault+0xa 0x2c3b980: 0xb585cb _ZN8fid_info16getFileEnum3InfoEP12_FILE_ INFO_3+ 0xee897970: 0xb58970 _ZN14smb_connection11srvsvcAuditERiP16cS rvsvcF 0xee897a10: 0xb5837d _ZN14smb_connection11srvsvcAuditERiP20cS rvsvcA 0xee897a30: 0xb581d2 _ZN13smb_streamCtx16srvsvcAuditParseEP20 cSrvsv 0xee897a70: 0xb57f8f _ZN10smb_server16srvsvcAuditParseEP20cSr vsvcAu

Fixed the race condition that occurred when an enumeration of open files is being performed at the same time as an open file request is failing.

419382 / 41240528

2 The Data Mover allowed the user to zero out (truncate) a file even though he did not have any rights to do so.

If there is no write permission from the ACL, the open now fails and the file is not truncated.

422369 / 41403684

148

Version 6.0.51.6



1 If a file system became full, when user cleared the windows log file, the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../sched.cxx at

line: 154 SYSTEM WATCHDOG. The stack trace showed: f3a0: 0x14dd29 system_watchdog+0x61 f3b0: 0x1407b3 t_nmi+0xc3 f468: 0xa653db _ZN11BasicVolume18completionCallbackEP10 ccb_hea f4a8: 0xae10b1 _ZN14ScsiBus_Device12sim_finishIOEP14Scs i_IOReq f4b8: 0xae1348 _ZN14ScsiBus_Device10sim_finishEP14Scsi_ IOReque f4d8: 0xaede04 _Z14FCP_completeIOPv+0xf2 f528: 0xb4db26 cpd_osw_finish_srb+0xb0 f588: 0xb46b8d fc_exm_exchange_complete+0xfa

This issue occurred because the memory state of the log object was not correctly restored. This has been fixed, and the Data Mover will no longer panic under this condition.

425818 / 42018528

1 While using FileMover with many files (in this case, 873), an Arkivio server returned an invalid 0xFFFF value for File ID, which led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: /disks/Archive_Area_nsg7/6_0_41_303/Dart/s erver/src/cifs/client/include/cifsclient_n ode.hxx at line: 281 : getting size of an

unopened file. The stack trace showed: f21400: 0x141439e _ZN17DHSM_CIFS_RemoteP15rem

This issue was fixed by changing the default value (for unitialized FID) from ushort 0xFFFF to ulong32 0xCC11FFFF.

426242 / 42082680

149

Version 6.0.51.6



oteBlockReadEjR 11 f21590: 0x13f374a _ZN10DHSM_Vnode15remoteBlockReadEjR11File_ St f21610: 0x13c7c67 _ZN10MGFS_Vnode14migrateToBlockEjRiPP13Buf fe f21890: 0x13ed849 _ZN10DHSM_Vnode14migrateToBlockEjRiPP13Buf fe f218e0: 0x141169b _ZN10DHSM_Vnode10blocksReadEjjRiPP13Buffer _C f219d0: 0x13f1c94 _ZN10DHSM_Vnode9blockReadEjR11File_Status+ 0x f21a00: 0x7bc897 _ZN9File_Fork17readBlockInternalEjRP13Buff er_ f21a40: 0x7c70dc _ZN13File_OpenFile16readDataInternalEyjR11 Fil

1 A page fault PANIC might occur when changing quota state to different values (ON/OFF) from both Windows and from Control Station side at the same time for the same file system.

A lock on quota objects was added to prevent this condition.

426890 / 42139428

1 When an SMB2 client opened a file under a Quota Tree, then disconnected just before the file system unmount processing started, a file system unmount did not complete, which led to a loss of access.

the unmount processing was modified to properly clean up SMB2 durable handles on a Quota Tree.

427655 / 42232932

150

Version 6.0.51.6



2 CIFS CEPP events were not generated or incorrectly generated. This caused some third party applications to not work properly.

Code was modified to better manage CIFS CEPP events.

386670

2 When user applied a strong security policy to his W2K8R2 CIFS domain, FileMover, CAVA or CEPA on this Microsoft machine may fail with ACCESS_DENIED errors due to communication errors with the domain controller.

Code now uses the correct CIFS-based SPN (service principal name) to access the domain controller.

425859 / 41925924

1 When CIFS client application used file range lock operations intensively with conflicts between different clients, a race condition led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0 Err code: 0 Target addr: 0 **. The stack trace showed: 0x2c3bbd0: 0x14124c waitForReboot+0x90 0x2c3bbf0: 0x1414d5 fault_dump+0x67 0x2c3bc10: 0x13a398 handle_fatal_fault+0xa 0x2c3bc50: 0x768586 _ZN14File_LocksData6umountEy+0xc0 0x27de1c30: 0x7681ef _Z12lockd_umounty+0x123 0x27de1c70: 0x72dafe _ZN15unmountCallBack7executeEy+0x20 0x27de1c90: 0x72e8a3 _ZN14VFS_FileSystem10unmountNowEi+0x4d 0x27de1cb0: 0x72ebb5 _ZN14VFS_FileSystem9doUnmountEv+0x109

This race condition between range lock and close operations has been eliminated.

429671 / 42298318

1 Concurrent command to start usermapper, particularly when the database was disabled, caused a Data Mover panic with the following

Modified code to protect Usermapper database initialization from concurrent start.

430768 / 42669876

151

Version 6.0.51.6



DART panic/fault message: *** GP exception. Virt ADDRESS: 0x00012c5c2a. Err code: 0 *** . The stack trace showed: 6760: 0xbf96c0 _ZN7DskDump9checkdumpEPKc+0x7a 67a0: 0xbf98ce dump_return+0x4 67b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory _Prot 67d0: 0x14ce5d fault_dump+0x89 67f0: 0x13378d handle_fault+0x1d a57b0: 0x12323c3 __db_c_close+0x35 a57f0: 0x123271d __db_c_close+0x38f a5840: 0x1233169 __db_c_get+0x7a3

2 Whenever a new user or group quota was created, it didn't show up in the appropriate Unisphere quota list until an internal refresh.

Code was fixed to ensure new user or group quotas are displayed when created.

378714

1 Unable to access CIFS share when trying to rename a file that was waiting to be checked by virus checker and all VC servers were offline.

Code now checks if the client has disconnected before renaming a file using CIFS.

414744 / 40668160

2 New users could not be mapped via LDAP.

If a current LDAP configuration is deleted and a new configuration is created, CIFS code now automatically switches to the new LDAP setup.

414760 / 40449194

152

Version 6.0.51.6



1 During a share enumeration, if the share name contained 118N characters that generated a NULL hash key, renaming the share could cause a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x00011ef9c4 Err code: 0 Target addr: 0x0000000000 **. The stack trace showed: 0760: 0xc1ac70 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xc1ae7e dump_return+0x4 07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory_Prot 07d0: 0x14ce5d fault_dump+0x89 07f0: 0x14162d handle_fault+0x1d cbb7d0: 0xc3e590 _ZN11smb_sdShare8getDBKeyEPKt+0x1a cbb820: 0xc37564 _ZN11smb_sdShare8getDBKeyEv+0x54 cbb840: 0xc377f3 _ZN9shareList9getLockedERP9smb_sharePKcP9xdrS

Shares having a NULL hash key are now handled correctly, and the Data Mover will no longer panic under this condition.

418870 / 41139864

2 Samba clients were unable to use Kerberos authentication. The following messages were found in server_log: SMB: 3: SSXAK=RPC_NT_INVALID_BINDING Client=a.b.c.d origin=550 stat=0x40000,0

They indicate that Kerberos checksum type cannot be checked due to usage of CKSUMTYPE_HMAC_MD5 checksum type instead of regular Windows client's CKSUMTYPE_KG_CB=0x8003. As a result, Kerberos authentication cannot complete, but client falls back to NTLM authentication seamlessly.

Code has been fixed to enable support for CKSUMTYPE_HMAC_MD5 checksum type.

427779 / 42247186

153

Version 6.0.51.6



1 Memory was not correctly released after a DIC sending error. This memory leak led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../bufpool.cxx at line: 1309 : couldn't get a free page. The stack trace showed: 08e90: 0x1ff58e _ZN18Buffer_Cached_Pool17makeCachedBuffersEm+0 08ee0: 0x14497f _ZN19Memory_AddressSpace16makeCachedBufferEv+0 08f10: 0x1ff631 _ZN18Buffer_Cached_Pool12findAfreeBufEmiii+0x6 08f50: 0x1ff6b9 _ZN18Buffer_Cached_Pool15getAnyCachedBufEmiii+ 08fa0: 0xa8f082 _ZN17UnCachedIO_Object9readBlockEyj16BlockIO_P 09060: 0x8ba443 _ZN14UFS_FileSystem9readBlockEiiRP13Buffer_Cac 090c0: 0x8b4550 _ZN9UFS_Vnode5breadEjR11File_Statusi+0x352 09130: 0x92a676 _ZN9UFS_Vnode9blkatoff2EyRPvRP13Buffer_Cachedi

Memory allocated for the issue is now released once it has been used.

432293 / 42848676

154

Version 6.0.51.6


Control Station


Severity Symptom Description Fix Summary Tracking Number 2 Administrators with a space

in their ID could not process Join/Unjoin requests. Celerra replaced spaces with an asterisk which rendered the admin ID invalid

Corrected code to ensure that spaces embedded inside 'admin' are handled correctly.

406140

2 Following a NAS code upgrade, when trying to query the status of SNMP on the Control Station, the system was returning the following error: nasadmin]# /sbin/service snmpd status snmpd dead but subsys locked /sbin/service snmpd stop Stopping snmpd: [FAILED]

Code has been fixed to save the old config files and properly update the new config files during upgrade.

409494 / 39688446

1 After a power outage the CIFS servers came back as NT4 type instead of Windows 2000 type.

Code has been fixed to check for an existing server entry with the same name as the new server entry. If found, the existing entry and the new entry get merged to create a new entry. If the new and existing entries differ in server type, or domain, or netbios (W2K only), code now replaces the new entry with the existing one.

381939 / 36515374

1 The Control Station ignored the event dispatch policies of the CLARiiON, causing a

The Control Station no longer ignores these event dispatch

415637 / 39816782

155

Version 6.0.51.6


Severity Symptom Description Fix Summary Tracking Number very high amount of call homes which led to the flooding of the call home utilities and servers.

policies.

CORE


Severity Symptom Description Fix Summary Tracking Number 2 When issuing the warm

reboot, if the Data Mover panicked for any reason, Data Mover failover occurred.

Code has been fixed to ignore any unexpected events during warm reboot PreShutdown operation and not fail over.

403054

Deduplication

The following entries describe fixed issues related to Deduplication.


1 The deletion of deduplicated files while open transactions were in progress led to a deadlock situation where the UFS transaction log was unable to make progress. DART panic/fault message: >>PANIC in file: ../ufsvfsop.cxx at line: 278 : sync: no progress reducing dirty

list. The stack trace showed: 0x2752bf40: 0x141278 waitForReboot+0x90 0x2752bf60: 0x141501 fault_dump+0x67 0x2752bf80: 0x1413e9 PANIC+0x29 0x2752bf90: 0x7dcb25 _ZN21UFS_SyncThreadInitial5startEv+0x17f 0x2752bfe0: 0x145c26

Sthread_startThread_internal+0xe

Changed the way file system deletes compressed data produced by deduplication to prevent this panic.

422705 / 41604572

156

Version 6.0.51.6



1 When using alternate data streams on a Deduplication-enabled file system, removing a deduped unnamed stream file caused the Data Mover to panic with the following DART panic/fault message: isAValidIndBlkBuf: bad bn. The stack trace showed: 0x21e47b40: 0x141280 waitForReboot+0x90 0x21e47b60: 0x141509 fault_dump+0x67 0x21e47b80: 0x1413f1 PANIC+0x29 0x21e47b90: 0x784dd7 _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0x21e47d50: 0x79b64a _ZN14UFS_FileSystem17isAValidIndBlkBufEP 13Buff 0x21e47dd0: 0x7915c3 _ZN9UFS_Vnode17readIndirectBlockElRP13Bu ffer_C 0x21e47e00: 0x791d25 _ZN9UFS_Vnode4bmapElRlPiS1_+0x237 0x21e47e60: 0x10366f0 _ZN10DHSM_Vnode14itruncInternalERP20UFS_ Update

Code has been fixed to ensure blocks are flushed to disk properly before files are removed or truncated.

432149 / 42837196

2 User reported poor performance and high CPU utilization during a deduplication scan on file systems which contained thousands of hard linked files.

Replaced linked list with hash table to significantly improve scan times on file systems containing thousands of hard linked files.

424654 / 41006630

1 User was unable to read deduplicated files. The following error was displayed in the server_log: corrupt mignode for offline inode x. This can happen if the stub file and migration inode uid / gid is out of sync.

This new code will allow the migration inode to be reingested even if the UID/

421654 / 41472302

157

Version 6.0.51.6



GID is out of sync so that stub file read operations are possible. It will also try to re-sync the uid / gid on file system update operations such as write, setLength, or recall.

1 Compressing files caused their ctime to be updated which triggered unnecessary activity for incremental backup even though neither the data nor the file meta-data had been changed from the users perspective.

Code was modified to ensure deduplication and re-duplication do not impact the ctime for either the internal Dart PE scan or for CIFS client compression / decompression.

424721

FileMover

The following entries describe fixed FileMover related problems:

Severity Symptom Description

Fix Summary Tracking Number

158

Version 6.0.51.6




1 While doing read operations using FileMover, threads hung and prevented CIFS access when the secondary storage became unavailable.

During read operations, code was modified to require a shared lock only under this condition which will prevent hung threads.

417842 / 41062586

1 Group quota usage is incorrect after recalling file from secondary file server.

This issue occurred because of a typo in DHSM feature that typed a GID as UID. This quota usage issue has been corrected.

440598

File Systems

The following entries describe fixed File System problems:


1 Write performance degraded substantially on a file system with checkpoints compared to a file system without checkpoints. Performance degradation was caused by inefficiencies in checkpoint handling code path.

The code has been optimized to improve write performance on file systems with checkpoints.

429333

1 Following an upgrade to version 6.0.42.0, when reading a deduplicated file for the first time and creating/refreshing a checkpoint, the checkpoint was marked as corrupted and unmounted. This was due to improper lock handling and creating a checkpoint with a dirty flag while there was a superblock update on the production file system.

Code has been fixed to properly use the lock and prevent checkpoint from being falsely marked corrupted under this condition.

429662 / 42529536

1 Migration between CFA and FMA was failing This issue occurred 382368 /

159

Version 6.0.51.6



for a file system with Conversion file is different than an existing one error observed in the server log.

because the file system was replicated from an older system while a conversion was in progress. This caused the new FS to be marked as "conversion active", preventing the new conversion from failing. Code has been fixed to properly handle this condition and finish the conversion on the new destination automatically.

36303236

2 The server_df server_x -inode command reported a negative a number of free inodes if the file system was created with lower than the default setting for nbpi (number of bytes per inode). This was because server_df was not able to report more than 2G free inodes.

The code has been fixed so that server_df is now able to display the correct number of free inodes regardless of the nbpi setting.

376435 / 35689866

160

Version 6.0.51.6



1 When accessing a large directory (in this case 1.25 GB) and storing hash of the directory into memory, code did not properly release the memory which led to a Data Mover panic with the following DART panic/fault message: >PANIC in file: ../bufpool.cxx at line: 421 : couldn't get a free page. The stack trace showed: 0x79d0f8c0: 0x140ce4 waitForReboot+0x90 0x79d0f8e0: 0x140f69 fault_dump+0x67 0x79d0f900: 0x140e51 PANIC+0x29 0x79d0f910: 0x1c6af9 _ZN18Buffer_Cached_Pool17makeCachedBuffersEi+0 0x79d0f940: 0x13c9c7 _ZN19Memory_AddressSpace16makeCachedBufferEv+0 0x79d0f960: 0x1c669c _ZN18Buffer_Cached_Pool15getAnyCachedBufEjii+0 0x79d0f990: 0x77e8eb _ZN14UFS_FileSystem8getBlockEliRPv+0x49

Code has been fixed to properly reclaim memory under this condition and prevent the panic.

383034 / 36705796

2 When user tried to rename a file without write access into target directory, the rename failed with "Access denied". When this occurred, some CEPP events were generated twice for some NFS V3 operations.

Code was modified and will only send the CEPP PostEvents once.

491155

2 NFS lock was not preserved upon a Data Mover failover.

Code now preserves the statd registry when file systems are unmounted by administrator.

407035 / 39509656

1 Some version of AIX clients request NFSv4 4.1 attributes with a 4.0 mount. This caused AIX 6.1 host to crash when trying to mount a file system using NFSv4.

If the client is 4.0 we return only 4.0 attributes even if 4.1 attributes are requested.

409247 / 39816612

161

Version 6.0.51.6



1 After running FSCK on a deduplicated file system, some files were inaccessible.

Improved FSCK handling of deduplicated files containing Alternate Data Streams (ADS).

410661 / 39762892

2 When using NFSv4, if a client sends a bad StateId this could result in NFSv4 errors, forcing the user to umount and remount NFSv4 file systems to get access again.

Code has been fixed to properly handle NFSv4 owner sequence and prevent this condition.

419732 / 40998286

1 Write performance degraded substantially on a file system with checkpoints compared to a file system without checkpoints. Performance degradation was caused by inefficiencies in checkpoint handling code path.

The code has been optimized to improve write performance on file systems with checkpoints.

429333

162

Version 6.0.51.6



1 When using NDMP, a file was truncated when backing up the R/W file system on which the deduplicated file resided. The last coming data size exceeded the bytes in the last block and caused a Data Mover to panic with the following DART panic/fault message: Memcpy size 0xfffff28e itoo large. The stack trace showed: 909a68: 0x25147e memcpyPanic+0x42 909ac8: 0x282578 memcpy+0x54 909b40: 0x140009a _ZN19File_OpenDedupeFile11writeStreamEPhjRj+ 909bc0: 0xbf32d9 _ZN15PaxDedupHandler11WriteStreamEyPcjRj+0xaf 909c10: 0xbf469d _ZN14pax_cmd_thread13dedup_restoreEP5ARCHDP15 909d50: 0xbafcc0 _ZN14pax_cmd_thread7extractEP5Event+0x31a0 909f40: 0xbb1004 _ZN15restore_Initial8appStartEv+0x24a 909fb0: 0xb46723 _ZN15Uthread_Initial5startEv+0xf

Code now compares the size of data it is trying to write with the remaining space of the migVnode (before EOF), and writes the smaller. If there's not enough remaining space, code returns EOF.

413977 / 40557024

1 When using NMFS and concurrent CIFS and NDMP access, a race condition occurred if two sharedLocks on the same thread had an exclusiveLock request.

The fix eliminates the second sharedLock attempt on the thread.

415992 / 40821538

1 Loss of access when attempting to generate a tree quota report while the tree quota directory was being deleted.

Changed the locking associated with looking up tree quota paths to eliminate a potential deadlock.

416425 / 40875472

1 When Restricted Groups contained a domain SID, the Data Mover panicked when applying GPO that contained Restricted Groups. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS:

The Data Mover now handles domain SID in Restricted Group correctly.

417912 / 41067196

163

Version 6.0.51.6



2020c9 Err code: 0 Target addr: 0 **. The stack trace showed: 0x2c4d92c: 0x14124c waitForReboot+0x90 0x2c4d94c: 0x1414d5 fault_dump+0x67 0x2c4d96c: 0x13a398 handle_fatal_fault+0xa 0x2c4d990: 0xc7c6fb _ZN17abstractConverter12strTranslateERPKcS1_+0x 0xe61fd9d0: 0xcbb8cc _ZN16localgroupServer11name2MemberEP13smb_thre 0xe61fdaf0: 0xcbbd3c _ZN16localgroupServer13names2MembersEP13smb_th 0xe61fdb20: 0xcbc99b _ZN16localgroupServer12addRGMembersEP13smb_thr 0xe61fdbd0: 0xcbc428 _ZN16localgroupServer19addRestrictedGroupsEv+0

2 Renaming a file from an NFSv4 client took an unnecessary lock on the parent directory as well as locking the file itself. This lock on the parent directory was not released when the rename was finished, and had to wait for 10 seconds before it timed out.

The code now correctly only locks the file being renamed, and not its parent directory.

418716 / 40017698

2 When user/group quota was enabled on a file system, the default file system limits were not inherited by the user/group quota records.

User/group quota limits are now set correctly.

420491 / 41347324

164

Version 6.0.51.6



3 Following an upgrade from Version 5.6 to 6.0 in a Linux environment, NFSv4 ACLs were not inherited from the parent and files created could not be used because the permissions were not correct.

When mode bits are set with NFSV4 in MIXED access policy, inheritance is preserved by a special ACL setting. Access right control on this processing was inconsistent with setAttr right checking. This has been fixed.

424387 / 40438280

2 If an ACL for a file or directory contained an entry for CIFS-only well-known groups, NFSv4 users were unable to modify the ACL of the file. This is blocked intentionally, to prevent NFSv4 clients from replacing the CIFS-only ACE with an invalid entry. However, there was no indication why the user couldn’t change the ACL.

When this occurs, the following message is now logged in the server_log: nfs4_checkSpecialA ce : SID <SID> cannot be translated to NFSV4 : ACL can't be modified.

425723 / 41929606

1 When user changed ownership for millions of files, a memory leak condition accumulated to the point where the Data Mover panicked with the following message: >>PANIC in file: ../addrspac.cxx at line: 529 : Out of memory. The stack trace showed: 0x56c7fb80: 0x146dfe _ZN19Memory_AddressSpace10allocPagesEl12Memory 0x56c7fbc0: 0x8e54ff _ZN9UFS_Vnode15blockChainWriteEP10BlockChainiR 0x56c7fce0: 0x8e6183 _ZN9UFS_Vnode15blockChainWriteEiP10BlockChain+ 0x56c7fd50: 0x150dd10 _ZN10DHSM_Vnode15blockChainWriteEiP10BlockCha

The memory leak issue has been fixed.

428118 / 42269136

165

Version 6.0.51.6



0x56c7fda0: 0x86b34c _ZN9VFS_Vnode18writeChainOfBlocksEP10BlockChai 0x56c7fde0: 0x7ed2c0 _ZN9File_Fork5flushERyii+0x1ac 0x56c7fe20: 0x7ed7ee _ZN9File_Fork16commit_cachedataEjji+0x158 0x56c7fe60: 0x7fc343 _ZN13File_OpenFile5closeEiii+0x55

1 User reported a higher than expected CPU usage during deduplication operations.

Eliminated unneeded debugging options which were found to be contributing to the CPU load during deduplication operations.

453242 / 43990760

166

Version 6.0.51.6



1 When using NFS V4, revoking an open state from an unconfirmed owner could double release the state, because in some cases the pointer to the (unreferenced) state was not reset to NULL. Since states are recycled, the release in excess caused a missing reference on a valid state. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 674dbb Err code: 0 Target addr: 0 **. The stack trace showed: 0x2c46464: 0x14124c waitForReboot+0x90 0x2c46484: 0x1414d5 fault_dump+0x67 0x2c464a4: 0x13a398 handle_fatal_fault+0xa 0x2c46580: 0x66e244 _ZN11NFSv4_Owner4openEmmP8nfs4_ctxP8stateid4RP 1 0xdffb9590: 0x6704e7 _ZN16NFSv4_StateTable13OpenClaimNullEmmP11NFSv 0xdffb9a10: 0x662b4b _Z9nfs4_openRP4msgbS1_RPhR8nfs4_ctx+0x8a9 0xdffb9d90: 0x65c870 _Z23nfs4_convertNfsv3Statusm+0x626 0xdffb9ee0: 0x63ebf1 _ZN17nfs_serverInitial5startEv+0xad7

This issue was fixed, and the state pointer in the owner is reset to NULL before releasing the state.

407115 / 39374052

1 When running an online quota check for a file system containing a quota record of anonymous ID, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x00008c7dc6 Err code: 2 Target addr: 0x000000005c **. The stack trace showed: 0760: 0xc1ac70 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xc1ae7e dump_return+0x4 07b0: 0x14c7d2

Code was fixed to ensure the quota record of anonymous UID no longer causes this panic during online quota checks.

421173 / 41453586

167

Version 6.0.51.6



_ZN12Memory_Frame13setProtectionEm19Memory_Pro t 07d0: 0x14ce5d fault_dump+0x89 07f0: 0x14162d handle_fault+0x1d f7a70: 0x8d72c6 _ZN14UFS_FileSystem14readQuotaDescsE9quotaType f7ee0: 0x8f0f9c _ZN14UFS_FileSystem9oqcEnableEP20onlineQuotaCh f7f30: 0x8f1221 _ZN20onlineQuotaCheckTask10initializeEv+0x149

2 When using NFSv4, if a file system was unmounted, some Linux clients could not return their delegations. Delegation are correctly recalled. But the client try an OPEN_DELEGATE_CUR to convert its delegation to an open and then flush its cache. This fails because the file-system is frozen.

During the umount process, code now allows OPEN CLAIM_DELEGATE_CUR from clients holding delegations to allow flushing the client cache

428074

2 When using NFSv4, client were unable to write to a file opened with a share reservation. This issue was due to I/O having write delegations conflict with share reservation

Code was fixed to ensure delegation states are global to a client and no longer conflict with other states

428236

1 When the NFS file system mounted on the Control Station became inaccessible, issuing the nas_version hung.

Code has been fixed to allow the nas_version command to work properly under this condition.

424679

168

Version 6.0.51.6


Installation/Configuration/Upgrades


Severity Symptom Description Fix Summary Tracking Number 1 During an upgrade, the nas_version

command hung if __db.00* lock files under /var/sadm/pkg/emcnas were corrupted.

Code now removes the lock files that are no longer in use, preventing the command from hanging due to prior corruption.

404601 / 3988124

1 During NAS upgrade, NAS service was required to be stopped on Control Station. In some cases when NAS service failed to stop, and Control Station reboot was initiated, the upgrade failed after reboot on the next step because pre-reboot task had not completed successfully.

With the fix, the upgrade will make sure that pre-reboot task is complete before continuing to the next step in upgrade.

417369 / 40921968

2 The upgrade Blade parameters task failed during out-of-family upgrades from 5.6 to 6.0 if the system had non-default host names configured for the Data Mover.

Code now correctly ascertains slot number when non-standard DM names are in use.

414362

2 User wanted to remove a portion from their emailuser alerts. There was a "System Configuration Snapshot" at the bottom of the e-mail that the user did not want to include. It included a "/nas/sbin/model" command and a "nas_inventory -list". The primary goal was to remove the "nas_inventory" command output from these e-mails.

A mechanism to control whether or not inventory data should be included into email notifications has been introduced.

423814 / 41763194

1 When booting a Control Station with ExpressInstall CD/DVD, if user selected Install option, the process did not verify existing backend data, and overwrote it. This was inconsistent with non-ExpressInstall CD/DVD.

Fixed the code so ExpressInstall CD/DVD work the same way as non-ExpressInstall

421316 / 41453962

169

Version 6.0.51.6


Severity Symptom Description Fix Summary Tracking Number CD/DVD.

2 When using CUT to upgrade to version 6.0.50, nas_version command displayed the following error: $ nas_version /nas/bin/nas_version: line 21: lsof: command not found rm: cannot remove `/var/sadm/pkg/emcnas/__db.001': Permission denied rm: cannot remove `/var/sadm/pkg/emcnas/__db.002': Permission denied rm: cannot remove `/var/sadm/pkg/emcnas/__db.003': Permission denied

Fixed command to properly handle this error condition and no longer report these errors.

421334

1 Celerra was not picking up failed drive events

The file that indicates events to the Celerra operating environment software was updated so that drive failed/removed messages are handled correctly.

432919

NDMP

The following entries describe fixed Network Data Management Protocol/Backup problems:



170

Version 6.0.51.6




1 When running NDMP, if the getinq command failed, it accepted the current value of a different buffer. If the different buffer had random data, it resulted in a malloc failure, which led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 515 : malloc() - incorrect size - not a positive value. The stack trace showed: 0x4edce14d0: 0x25d13e malloc+0x5e 0x4edce1520: 0x1723c25 _Z18GetScsiDevicesNameRP16scsi_device_namei+ 0x4edce17e0: 0x1724283 _ZN24NdmpdConfigGetTapeInfoMH14messageHandle 0x4edce1880: 0x172c669 _ZN14NdmpConnection15processMessagesEi+0x509 0x4edce1910: 0x172c84f _ZN14NdmpConnection15processRequestsEv+0x9 0x4edce1920: 0x16f836d _ZN21ConnectionFileHandler11handleEventEv+0x 0x4edce1940: 0x172d38f

The NDMP code now checks the status of the getinq function and exits properly, if the getinq command fails.

430903 / 42684894

171

Version 6.0.51.6




_ZN19NdmpEventDispatcher8dispatchEij+0x729 0x4edce1a80: 0x16fb29a _ZN21NdmpConnectionHandler17connectionHandle

172

Version 6.0.51.6




1 While performing a VBB restore, during file creation, the lookupPath() failed. If the func call did not find the path, it returned NULL. The Data Mover panicked with the following DART panic/fault message: *** DART panic/fault message: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 124a3c0 Err code: 0 Target addr: 0 ** The stack trace showed: 0x2c62e14: 0x141278 waitForReboot+0x90 0x2c62e34: 0x141501 fault_dump+0x67 0x2c62e54: 0x13a398 handle_fatal_fault+0xa 0x2c62e90: 0x1246211 _ZN14vbbCreateFiles21createDestinationPathEPcm 0x58545f20: 0x1245c3d _ZN14vbbCreateFiles14createBaseTreeEv+0x463 0x58545fa0: 0x1244d84 _ZN14vbbCreateFi

The panic occurred when VBB used the returned NULL pointer. The fix checks the return code and returns failure if the function call fails, instead of panicking the Data Mover.

416911 / 40908030

173

Version 6.0.51.6




les3runEv+0xde 0x58545fc0: 0x125422b _ZN24vbbProjectMember_Initial5startEv+0x1f 0x58545fe0: 0x145c26 Sthread_startThread_internal+0xe

1 If the residual data and actual read data of a file during an NDMP restore did not agree, the restore failed because the data read from tape was not what was expected. ERROR MESSAGE: Tape Silvering restore with error of extent length not multiple of 8K.

This tape read error was fixed. Code now ignores residual data if the read returns OK.

409887 / 39959654

2 While doing a VBB DAR full restore, the attributes of directories were not reset to the values saved in the archive. This issue did not exist in non-DAR restore.

The attributes of directories are now reset to the values saved in the archive.

411250

174

Version 6.0.51.6




1 When backing up a file system to tape, a media error caused the server_archive command to fail and the Data Mover to panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000ca204c Err code: 0 Target addr: 0x0000000004 **. The stack trace showed: 08e0: 0xce0520 _ZN7DskDump9checkdumpEPKc+0x7a 0920: 0xce072e dump_return+0x4 0930: 0x14d9f2 record_panic_ticks+0x242 0950: 0x14e07d fault_dump+0x89 0970: 0x141da5 handle_fault+0x1d 6ebd30: 0xca26ee _ZN14pax_cmd_thread9wr_rdfileEP5ARCHDiP11MsgL 6ebdd0: 0xc99f78 _ZN14pax_cmd_thread10wr_archiveEv+0x1f42 6ebf40: 0xc9afbb _ZN14backup_Initial8appStartEv+0x243

NULL pointer protection was enhanced to prevent this panic.

422966

2 When using NDMP, during reverse block level object (BLO) recovery, the cmd

The reverse BLO (block level object) recovery code was modified to work

424158 / 41813810

175

Version 6.0.51.6




service could not find the task context and logged the following error to the server log every 5 minutes - filling up the server log : 2011-06-26 04:07:23: CMD: 3: handleTaskObject: INDICATION action TASK Context not found for Task = APM001110019772007_2_53:0 2011-07-01 12:23:18: CMD: 3: handleTaskObject: INDICATION action TASK Context not found for Task = APM001110019772007_2_53:0

correctly and no longer log these errors to the server_log.

1 After using NDMP/PAX to backup deduplicated files, and then attempting to restore them on a new file system (or a different file system that did not have dedupe enabled), user saw the following errors in the server_log... 011-10-28 18:23:44: DEDUPE: 3: builder: RDE not initialized fsid 178 2011-10-28 18:23:44: DEDUPE: 3: writeStream: failed to get ssp and none of the deduplicated files will be restored on that file system.

Fixed the deduplication initialization for this case.

435544

176

Version 6.0.51.6




1 While using NDMP, user was unable to run 8 simultaneous server_archive sessions, but could only get 4 to run in parallel, even on 8gb memory systems.

Code was fixed to enable user to run 8 server_archive on 8gb memory systems.

429341

1 NDMP failed to restore files with error message SHADOW: 6: setName: Cannot translate xxx in unicode (5=InvalidArgument) but did not return an error to the NDMP server

Code now sends this warning message to FMA/DMA.

424257

2 When user tried to do a DUMP backup and set file or directory filters, if the filters matched a symbolic link and the symbolic link could not be stat()-ed, the backup job failed. Here are some cases when stat() will fail: 1. destination of the symbolic link doesn't exist. 2. the depth of the symbolic link is higher than 8(File_MaxSymLinksInPath).

Code was changed to correctly account for symbolic links.

423990

177

Version 6.0.51.6


Network

The following entries describe fixed Network-related problems

Severity Symptom Description Fix Summary

Tracking Number

2 FTP behavior was not working as expected, the -a option on the ftp ls command did not return . and .. This caused an infinite recursive list loop. Eventually, the list command was truncated, resulting in 'command not understood'.

Updated the FTP server in order to allow .and.. to be shown in ftp ls -a command

412323 / 40150774

1 If the peer server disconnected while the HTTP client completed the initialization of the TCP connection, a race condition led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000213187 Err code: 0 Target addr: 0x0000000000 **. The stack trace showed: 0760: 0xc1ac70 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xc1ae7e dump_return+0x4 07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory_Pro t 07d0: 0x14ce5d fault_dump+0x89 07f0: 0x14162d handle_fault+0x1d 280d30: 0x41c2c0 readmsg+0xe0 280db0: 0x41c324 GetsocknameStream+0xe 280dd0: 0x10c064a _ZN14http_clientCnx11_tcpConnectEv+0xd84

This race condition has been eliminated.

421919 / 41579878

178

Version 6.0.51.6



Tracking Number

1 When there were lots of fragments (typically UDP traffic) and empty network buffers at the head of the fragment chain, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x00003ce5e0 Err code: 0 Target addr: 0x0000000000 ** . The stack trace showed: 0760: 0xc1ad80 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xc1af8e dump_return+0x4 07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory_Pro t 07d0: 0x14ce5d fault_dump+0x89 07f0: 0x14162d handle_fault+0x1d ed90: 0x3c3897 ip_mxrput+0x10d edd0: 0x37d8cd _Z7bcmreadPvP4msgbtjj+0x4df eee0: 0x38e1c2 _Z20LM_ServiceInterruptsP16_LM_DEVICE_BLOCK+0x 3

Code no longer panics when empty blocks are removed via MSG_TRIM().

422980 / 41687604

1 If a user made an SNMP bulk request with a repeat value of "0", the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 378 : malloc() - incorrect size - not a positive value. The stack trace showed: 825ce0: 0x24a290 malloc+0x5c 825d20: 0x10e49ef _Z28netsnmp_create_subtree_cacheP23netsnmp_a 825d70: 0x10e4eb0 _Z10handle_pduP23netsnmp_agent_session_s+0xc 825da0: 0x10e5d3e _Z22netsnmp_handle_requestP23netsnmp_agent_s 825dd0: 0x10e67c6 _Z18handle_snmp_packetiP12snmp_sessioniP8snm 825df0: 0x1116a2d _Z12snmpv3_parseP8snmp_pduPhPmPS1_P12snmp_se 825e40: 0x111868f _Z10_sess_readPvP8LinkListR12SelectStream+0x 825eb0: 0x11187de _Z14snmp_sess_readPvP8LinkListR12SelectStrea

Code now checks for non-null 'r' repeat values before allocating memory using 'r' value.

426298 / 42091892

1 When a BCM NIC was hung, it was immediately Changes 434044

179

Version 6.0.51.6



Tracking Number

reset.

were made to allow specifying that the BCM NIC device not be reset when it appears to be hung. This only impacts some troubleshooting scenarios.

1 Heavy usage of the TFTP protocol on a Data Mover caused memory overflow and led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x000089863e Err code: 3 Target addr: 0x0000b42f72 **. The stack trace showed: 1760: 0xbffe60 _ZN7DskDump9checkdumpEPKc+0x7a 17a0: 0xc0006e dump_return+0x4 17b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory_Pro t 17d0: 0x14ce5d fault_dump+0x89 17f0: 0x13378d handle_fault+0x1d 2bdd00: 0x8b9261 _ZN14UFS_CgBlkEntry11postFreeBlkEPvS0_+0x5d1

Code has been fixed to properly allocate/release the required amount of memory used by TFTP stack, and prevent this Data Mover panic.

434697 / 41576516

180

Version 6.0.51.6



Tracking Number

1 When using a Domain name in an export list for NFS client, if the DNS server was unavailable, a Data Mover panicked while the NFS client was trying to mount the export. DART panic message: *** Page Fault Interrupt. Virt ADDRESS: 0x00004a4c77 Err code: 0 Target addr: 0x0000000048 **. The stack trace showed: 0x36d0760: 0xc1ad80 _ZN7DskDump9checkdumpEPKc+0x7a 0x36d07a0: 0xc1af8e dump_return+0x4 0x36d07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memory_Prot 0x36d07d0: 0x14ce5d fault_dump+0x89 0x36d07f0: 0x14162d handle_fault+0x1d 0x834ecb0: 0x40db4e udp_input+0x9fa 0x834ee00: 0x3ce997 ip_input+0x1425 0x834eeb0: 0x3c3897 ip_mxrput+0x10d

If the DNS server is unavailable, code will no longer panic under this condition, but will post the following message to the server_log: Portmap session is already removed because of timeout.

437344 / 43462506

181

Version 6.0.51.6


Replication

The following replication-related bugs that were fixed in this release:



2 User could not create a checkpoint for a nested mounted file system for NDMP backup

The fix will allow checkpoint creation from nested mounted file system during NDMP backup.

407727

2 Unisphere did not display current transfer information of a replication session if the size was greater than 2TB.

Unisphere now displays the transfer information of a replication session regardless of the size.

413467 / 40439600

182

Version 6.0.51.6


1 If an ongoing replication copy was disturbed (for example: unmounting the base line checkpoint during a nas_copy session), the Data Mover incorrectly marked the copy as complete although there was an error during the copy. The copied file system remained inconsistent which caused the Data Mover to panic when the file system was accessed. DART/panic fault message: mangled directory entry. The stack trace showed: 0x109412b00: 0x89c445 _ZN14UFS_FileSystem28markCorruptedFsForU nmoun 0x109412c70: 0x928411 _ZN9UFS_Vnode6dirbadEjPKc+0xb1 0x109412d10: 0x90cb57 _ZN20UFS_DirHashInfo_dir3C1EP11Local_Vno de+0x 0x109412db0: 0x93e582 _ZN15UFS_DirOps_dir311initDirHashEP9UFS_ Vnode 0x109412df0: 0x933dc9 _ZN15UFS_DirOps_dir37scandirEP9UFS_Vnode PK18F 0x109412ec0:

Code now properly checks the status codes for local replication using the loopback interface. This prevents marking a replication session to be completed when one of the base line checkpoints is unmounted, and eliminates the panic.

418323 / 41087088

183

Version 6.0.51.6


0x91b353 _ZN9UFS_Vnode7scandirEPK18File_NameCompo nentR 0x109412ed0: 0x93083c _ZN9UFS_Vnode10lookupCompEPK18File_NameC ompon 0x109413000: 0x13bd49d _ZN10MGFS_Vnode10lookupCompEPK18File_Nam eCom

184

Version 6.0.51.6


1 When a checkpoint was taken, if the memory was fragmented and there was a lot of DBMS activity, the Data Mover panicked with the following .DART panic/fault message:>>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory. The stack trace showed: 0x5a9a1c30: 0x145966 _ZN19Memory_AddressSpace10allocPagesEl12 Memory 0x5a9a1c70: 0x24a34e malloc+0x11a 0x5a9a1cb0: 0x24ad2f realloc+0x81 0x5a9a1ce0: 0x1285087 __os_realloc+0x3c 0x5a9a1d10: 0x1283f76 __memp_sync_int+0x280 0x5a9a1de0: 0x1284822 __memp_sync+0xe6 0x5a9a1e10: 0x12902db __txn_checkpoint+0x1be 0x5a9a1eb0: 0x1290555 __txn_checkpoint_pp+0xf9

The fix consists of allocating all the memory when the environment is initialized.

421132 / 41400444

1 Too many duplicated final response for

Added a defensive mechanism to

398657 / 38323670

185

Version 6.0.51.6


DicEvent were queued in the cmd queue, and each of them consumed about 24KBytes which exhausted the Data Mover memory. DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 289 : Out of memory. The stack trace showed: 0xaae2dda0: 0x14124c waitForReboot+0x90 0xaae2ddc0: 0x1414d5 fault_dump+0x67 0xaae2dde0: 0x1413bd PANIC+0x29 0xaae2ddf0: 0x13bef6 _ZN19Memory_AddressSpace10allocPagesEi12Me mory 0xaae2de10: 0x10a0283 _ZN10DpBuf_Pool8getChunkEiiP16DpBuf_PoolMe mbe 0xaae2de30: 0x10a0386 _ZN10DpBuf_Pool8allocateEv+0x5c 0xaae2de50: 0x107fef8 _ZN9DpRequestC1E15DpRequestMode_e17DpReque stC 0xaae2de70: 0x10b1776 _ZN21DpService_ReplicaInfo19processInfoPar all

eliminate final response duplication.

186

Version 6.0.51.6


1 In a case where an error occurred during a tape silvering restore process, and the restore was aborted, a RedirectReads(FALSE) call in the code caused the file system to enter a freeze state and be set inactive. Restore from the internal ckpt then failed due to the inactive file system state.

The code was fixed by moving RedirectReads(FALSE) to the end of the tape silver restore, preventing this issue and allowing the restore to complete.

411982 / 39959654

1 The replication delete operation failed to delete the mount option record in the database. Leftover replication file system mount option record may silently change the file system read-write attribute after a reboot, even without replication on it.

Code now checks whether there is replication on the file system, and only saves the mount option record when necessary. In addition, replication code removes duplicated entries during replication start.

421209 / 41475440

Security

The following entries describe fixed security-related issues



2 When the LDAP Service was reconfigured with no bind DN or password specified, the LDAP connection (and associated 'ldap' thread) were 'orphaned' or no longer used by the LDAP Service. While the connection and thread are usually

If the connections are not used (idle for a period of time), they will be disconnected by the LDAP Server.

426901 / 42075838

187

Version 6.0.51.6




cleaned up when (idle) timed out at the LDAP Server, if an LDAP disconnected, and multiple reconfigurations were attempted, the number of connections rapidly grew, consuming memory, threads, and other resources.

2 The following known security vulnerabilities existed prior to this release: Apache HTTP server (CVE-2011-3192) Bzip2 (CVE-2010-0405), DHCP (CVE-2011-0997) GnuC libraries ( CVE-2011-1071 ) GnuTLS (CVE-2008-1948, CVE-2008-1949, CVE-2008-1950 ) JRE (multiple vulnerabilities) Libpng (CVE-2008-6218) LibXML (CVE-2009-2414 and CVE-2009-2416) Logwatch (CVE-2011-1018) OpenSSL (CVE-2011-3207, CVE-2011-3210, CVE-2011-0014),

Several packages were updated to address these known security vulnerabilities. JRE update includes Oracle Java SE Critical Patch Update June 2011.

392895 , 413785, 428931, 428932, 428933, 431357, 433301, 433470, 433475, 433477, 432273 / 42301928 42301296 42301876 42302062

188

Version 6.0.51.6


Storage

The following entries describe fixed storage-related issues


1 An environmental issue occurred with HBA 1, during a self-test, which led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG. The stack trace showed: cee0: 0x14dd29 system_watchdog+0x61 cef0: 0x132913 t_nmi+0xc3 cfd0: 0x155263 Sthread_startThread_internal+0x7

Code has been fixed to prevent this environmental issue from panicking the Data Mover.

408642 / 39774566

1 During a FLARE upgrade, when the LUN trespassed from the primary to the secondary processor, the Data Mover panicked. DART panic/fault message: >>PANIC in file: ../BVolumeIrp.cxx at line: 405 : IO failure on Vol:89, blkNo:42139696,

SKey:0x54. The stack trace showed: fed0: 0xa9fad5 _ZN8SgIO_IRP9irpFailedE14BlockIO_Statusy +0x477 ff20: 0xb3baa1 _ZN14CamDisk_Device8completeEP10ccb_head er+0x14 ff90: 0xabf9c4 _ZN8Scsi_Xpt16processCallbacksEv+0xea ffc0: 0xabfa3d _ZN18Scsi_serverInitial5startEv+0xd ffd0: 0x155263 Sthread_startThread_internal+0x7

Fixed a timing window where two threads could each detect that a trespass is required and attempted to trespass to the wrong SP. Code now locks the volume and tests whether the current SP has changed before switching paths and issuing the trespass request.

423378 / 41629776

189

Version 6.0.43.1


Version 6.0.43.1

Deduplication

The following entries describe fixed issues related to Deduplication.



1 Following an upgrade to version 6.0.42.0, when reading a deduplicated file for the first time and creating/refreshing a checkpoint, the checkpoint was marked as corrupted and unmounted. This was due to improper lock handling and creating a checkpoint with a dirty flag while there was a superblock update on the production file system.

Code has been fixed to properly use the lock and prevent checkpoint from being falsely marked corrupted under this condition.

429662 / 42529536

Version 6.0.43.0

File Systems

The following entries describe fixed issues related to File Systems:



190

Version 6.0.42.3


1 Celerra Network Server version 6.0.42 added a check for invalid Alternate Data Stream names. When a client tried to query an invalid streamname, the operation was rejected as expected but in the process a lock was left on the file or directory. This led to other threads getting blocked if they needed to gain exclusive access to the object.

The checking for invalid streamnames has been removed.

393688

Version 6.0.42.3


The following are fixed problems related to the Celerra Management Applications, i.e., Celerra Monitor, Celerra Manager, WebUI, or Celerra MMC Snap-ins:



2 File system names containing a space could not display correct space information on GUI. The usage bar was empty, indicating there was no space used, even if the file system had consumed some space.

Modified code to correctly return the correct space information for file systems with names containing a space.

35613 / 33731506

2 When a quota crossed the soft limit and the grace period was unlimited, that quota property was

Code now translates the unlimited value to string "No limit".

401502 / 38626372

191

Version 6.0.42.3


not displayed in the GUI. Code represented an unlimited grace period as -1, and translated that to a string "NoLimit". The code path for the above scenario did not translate the value to a string and it caused an exception trace.

CIFS and NFS

This release addresses a potential vulnerability which could, under certain conditions, be exploited to cause a Denial of Service (DOS) attack.

CIFS



192

Version 6.0.42.3


1 When using emcacl to set the security descriptor on a symbolic link on a Data Mover with the param cifs.acl.extacl=0x28 and CEPA enabled, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: c5f858 Err code: 0 Target addr: 1c ** . The stack trace showed: 0x2c39cb4: 0x14124c waitForReboot+0x90 0x2c39cd4: 0x1414d5 fault_dump+0x67 0x2c39cf4: 0x13a398 handle_fatal_fault+0xa 0x2c39d40: 0xc5228f _ZN8AppLibNT5closeEP8fid_infoiR11File_St atusP17 0xe4307ec0: 0xae5301 _ZN14smb1_threadCtx10replyCloseEP4msgb+0 x289 0xe4307f40: 0xbe4b0d _ZN14smb1_threadCtx7processEP4msgbi+0xcd f 0xe4307fa0: 0xbd92ec _ZN13smb_threadCtx5startEv+0x6d0 0xe4307fe0: 0x145bfa Sthread_startThread_internal+0xe

The Data Mover panics when trying to get the size of the file when setting the security descriptor if the file is a symbolic link. The code now checks if the target is a symbolic link before attempting to get the file size.

416553 / 40899086

1 A 0% CPU idle condition might be observed on CIFS servers with many existing connections.

Code no longer browses existing connections during authentication.

404202 / 38611382

1 If the computer name of the CIFS server differed from its NetBIOS name, an attempt to join a CIFS Server to a Domain from GUI failed with the error "compname <comp_name> does not exists in CIFS

Code was incorrectly using the netBIOS name instead of computer name. This has been fixed.

393863

193

Version 6.0.42.3


configuration.

2 When using Virus Checker with a file system mounted on an NMFS path, the path sent to the virus checker was incomplete. This resulted in the file not being scanned

When a CIFS user connects to a share attached to an NMFS, code now puts the mount path of all the file systems mounted on this NMFS in memory to ensure all files are scanned.

410012 / 39861830

1 CEPA service was intermittently disconnected when retrytimeout was set to a value less than 200ms. This issue occurred because the TCP heartbeat request time out is always 300ms or greater.

Code now forces a retrytimeout internally to at least 300ms for the heartbeat.

397281 / 37879634

2 CIFS clients saw CIFS request timeouts in environments with high latency and high utilization. If the TCP queue for a CIFS was full, typically because a client makes several parallel requests, or there is high network latency (such as with a WAN), then clients may see the Celerra appear to ignore some CIFS requests. If the queue remained nearly full, eventually clients could see a CIFS timeout. This could also happen with CIFS clients that were not acknowledging TCP data quickly, but making CIFS requests aggressively.

When the CIFS TCP queue was full, new responses were placed at the head of the queue instead of at the end. If the TCP queue never fully emptied (because there was always data outstanding on the network), then some CIFS responses might never be transmitted, leading to timeouts. The problem is resolved by appending new replies to the end of the TCP queue instead of the beginning.

391540 / 36829842

1 When running clients which don't fully respect protocol, our CIFS server became unavailable after consuming all internal threads resources due to pending requests.

Code now drops incomplete requests, thus avoiding blocked threads.

404372 / 39134760

194

Version 6.0.42.3


1 The Celerra CIFS client was unable to connect to CIFS servers using NTLMv2. This would mean that if Kerberos authentication was unavailable and NTLMv1 was disallowed, the Celerra would fail to open CIFS connections. This could cause failures for DHSM recalls/file access, for CDMS, FTP authentication with Active Directory, or CIFS authentication of users.

Fixed a DNS comparison issue preventing CIFS clients from using NTLMv2 in Active Directory environments.

399722 / 38523654

1 During a CDMS migration between 2 Celerras, the migration hung. On the target Celerra, user was getting various errors indicating, that the source wouldn't deliver the data. On the source, user was getting the following error at the exact same time at the fault occurred: 2010-12-11 00:50:42: SMB: 3:[codan] Too many open policies on pipe=cMsrpc_Policy for client=10.194.116.100 (257/256). During the migration, the LSA function opens policies, but was failing to close them properly, eventually leading to the process hanging.

Added a use function to close LSA or SAMR open handles and prevent this error.

398486

1 If a security descriptor (SD) was not defined, a memory leak condition occurred in shareEnum502 which led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../bufpool.cxx at line: 1309 : couldn't get a free page. The stack trace showed: 0x11757f080: 0x241633 _ZN18Buffer_Cached_Pool17makeCachedBuffe rsEm+ 0x11757f0d0: 0x14e514 _ZN19Memory_AddressSpace16makeCachedBuff

Prior to the fix, code was building a SD which was automatically deleted. With this fix, the SD is no longer deleted.

393688

195

Version 6.0.42.3


erEv+ 0x11757f110: 0x2416ef _ZN18Buffer_Cached_Pool12findAfreeBufEmi ii+0x 0x11757f150: 0x2417d7 _ZN18Buffer_Cached_Pool15getAnyCachedBuf Emiii 0x11757f1a0: 0xcf5d8b _ZN17UnCachedIO_Object9readBlockEyj16Blo ckIO_ 0x11757f290: 0xa6fc23 _ZN14UFS_FileSystem9readBlockEiiRP13Buff er_Ca 0x11757f300: 0xa647fe _ZN9UFS_Vnode5breadEjR11File_Statusi+0x4 de 0x11757f370: 0xb135f0 _ZN9UFS_Vnode9blkatoff2EyRPvRP13Buffer_C ached

1 When using a Windows backup tool to backup Unix files and data, a Data Mover became unresponsive if extended ACLs are validated.

Removed a condition where two shared locks were taken in sequence which led to a deadlock.

403587 / 38982154

196

Version 6.0.42.3


1 When an NFS connection using Kerberos for authentication failed, a function was called using the wrong variable as a parameter, resulting in a Data Mover panic when it tried to obtain data that did not exist. DART panic/fault message: *** GP exception. Virt ADDRESS: 0x0000732c16. Err code: 0 ***. The stack trace showed: 1760: 0xbf6470 _ZN7DskDump9checkdumpEPKc+0x7a 17a0: 0xbf667e dump_return+0x4 17b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 17d0: 0x14ce5d fault_dump+0x89 17f0: 0x13378d handle_fault+0x1d 38be40: 0x733648 _ZN10NfsGssInit12processReplyERP4msgb+0x 42 38be70: 0x733105 _ZN13NfsGssRequest8dispatchEP4msgbPvP17A ccess 38bf00: 0x6f443a _ZN17nfs_serverInitial5startEv+0xa5c

Code now uses the correct variable in the function call.

409927 / 39986788

197

Version 6.0.42.3


1 Data Mover panicked when a Windows 7 client connected to a CIFS share through a Riverbed WAN accelerator with the SMB2 protocol enabled. The Riverbed WAN accelerator sent an SMB2_IOCTRL request compounded with SMB2 commands which the Data Mover did not decode correctly. DART panic/fault message: >>PANIC in file: /disks/NSG_Maint8/6.0_ build _nfs_env/Dart/server/src/kernel/include/st ream.hxx at line: 1119 : datab::freeRef: db_ref <= 0. The stack trace showed: 245ea0: 0xdcc4fd _ZN14smb2_threadCtx7processEP4msgbi+0xf57 245f20: 0xdee8d6 _ZN13smb_threadCtx5startEv+0xc4c 245fd0: 0x155263 Sthread_startThread_internal+0x7

The Data Mover now properly handles compound SMB2 requests received in this manner.

411721 / 40195472

198

Version 6.0.42.3


1 During a GPO update, if the connection to the domain controller is lost, an error case was used that deleted an object without setting the associated pointer to NULL. At the end of the update, the pointer was checked and deleted if not NULL. This resulted in a double-free, which caused a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 525 : Memory corrupted. The stack trace showed: 06d720: 0x24aaaf free+0x79 06d760: 0x24ae7d _ZdlPv+0x9 06d770: 0xf2b7fc _ZN9DomainGpo18updateVdmGpoServerERK6Str ingi+ 06dd40: 0xf2be4b _ZN9DomainGpo15updateGpoServerERK6String i+0x4 06dd80: 0xf250ae _ZN4gpod5startEv+0x117e 06dfd0: 0x155263 Sthread_startThread_internal+0x7

Code has been fixed to set the pointer to NULL after the object is deleted to prevent this panic.

409288 / 39850746

2 When using third party VPN software to connect to a CIFS server from a client that had the same machine name as the domain name, NTLM authentication failed with LOGON_FAILURE error and users were not able to connect to CIFS shares.

The code has been fixed to ignore client names that match the domain name of the CIFS server, allowing authentication to succeed.

407582 / 39550232

1 A Data Mover panicked when trying to connect to a share on a file system which was corrupted. DART panic/fault message: ** Page Fault

Code has been fixed to check the state of the file system before processing the share request, thus

401537 / 38704210

199

Version 6.0.42.3


Interrupt. Virt ADDRESS: 72735b Err code: 0 Target addr: 1e8 ** . The stack trace showed: 0x2c4c914: 0x14124c waitForReboot+0x90 0x2c4c934: 0x1414d5 fault_dump+0x67 0x2c4c954: 0x13a398 handle_fatal_fault+0xa 0x2c4c9f0: 0x83e738 _ZN26passthroughShadowDirectory16lookupN ameLock 0x28a2fa40: 0x83f662 _ZN26passthroughShadowDirectory23lookupN ameLoc 0x28a2fa70: 0x836609 _ZN26passthroughShadowDirectory23interna lLooku 0x28a2fb00: 0x8255a6 _ZN19ShadowDirectoryBase15lookupComponen tEPtmP 0x28a2fb30: 0xc6b217 _Z15getnodefromRootP15File_NamingNodePKc RS0_+0

preventing this panic.

1 When there were many CIFS users accessing a Data Mover, the Data Mover might panic due to 64-bit higher memory footprint. DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory.

Code has been fixed to reduce the size of various CIFS objects to improve memory usage.

404117 / 39062050

200

Version 6.0.42.3


1 Under a heavy CIFS load with many antivirus threads, there were 94 Write I/Os in progress which caused a memory fragmentation issue led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory. The stack trace showed: 06b0: 0x145966 _ZN19Memory_AddressSpace10allocPagesEl12 Memory_ 06f0: 0x207f2e _ZN20Buffer_Uncached_Pool10addBuffersEji j+ 0xa60750: 0x2098fd _ZN20Buffer_Uncached_Pool14getUncachedBu fE15Str 0790: 0x20ae0d allocb+0x67 07c0: 0x20bc47 pullupmsg+0x79 0810: 0x231545 _Z10msgdsize_cPP4msgb+0x1a7 0860: 0x231770 dlpiWPut+0xea 08d0: 0x3d0276 ip_output+0x432

Implemented changes to the NIC driver to reduce the odds of panicking due to memory fragmentation.

400441 / 38611382

2 With the dfsutil CLI from windows 2008R2, user could not reset the SD of a DFS link.

Code has been fixed to properly reset SD on the DFS link using dfsutil.

403696

Control Station

The following entries describe fixed problems related to the Celerra Control Station.

201

Version 6.0.42.3




1 The MirrorView activate command caused Control Station to reboot twice, 1st time after adding NBS service on RDF7 to DM2, and 2nd time after adding NBS service on RDF10 to DM2. The command finally succeeded on the 3rd attempt.

Both RecoverPoint and MirrorView DR features have been fixed and prevents Control Station rebppts during the DR failover at remote NBS volumes setup.

388624

1 While using Unisphere and trying to enable user and group quotas in one step, only the user quota was enabled. The group quota enable operation failed.

Code hass been fixed to detect the enabling of both user and group quota and use the "-both" command line operation.

404960

2 LDAP connection in Unisphere under Settings->Manage LDAP Domain fails with the error 13693354039: CST (Common Security Toolkit) library internal error.

Code has been fixed to provide more information in the recommended action when this error occurs so that the user can correct it and proceed.

401078 / 38622682

Deduplication

The following entries describe fixed issues pertaining to Celerra Deduplication:


Tracking Number

202

Version 6.0.42.3



Tracking Number

2 If the remote file system ID was a substring of another file system ID, user could not take a nearcopy/farcopy backup snap. This issue occurred if the remote file system size query substring match returned more than one item.

Changed a remote query from sub-string match to exact match.

402256

2 If the file system had quota enabled with filesize policy, and there's sparse recalled files in the Deduplication feature, then after an online filesystem quota check finished, the quota block usage report will be incorrect.

The online quota check scanned both stub file and migration file, so the file size was calculated twice. This fix makes the online quota check skip the filesize of stub file if it has a sparse recalled migration file.

406699

2 If the group quota usage exceeded 4TB, and F-RDE was enabled, migrateFile failed. Only an internal error was logged, there was no specific error information in server_log to help issue diagnosis.

There is a maximum of 4TB of deduped data. Code will now log a more specific error information in the server_log.

403974 / 38996546

1 When performing NDMP restore of a deduped file system on a highly utilized Data Mover, a buffer overflowed and a subsequential Data Mover panicked with the following DART panic/fault message: *** GP exception. Virt ADDRESS: 0x0000964d88. Err code: 0 ***. The stack trace showed: 0760: 0xbf6490 _ZN7DskDump9checkdumpEPKc+0x7a 07a0: 0xbf669e dump_return+0x4 07b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot

Code has been fixed to properly declare all the functions and prevent this condition.

401349 / 38640140

203

Version 6.0.42.3



Tracking Number

07d0: 0x14ce5d fault_dump+0x89 07f0: 0x13378d handle_fault+0x1d b7a00: 0x13b584e _ZN10DHSM_Vnode30convertOfflineBlockChai nToHo b7a80: 0x13faf8c _ZN19File_OpenDedupeFile10writeHolesEjj+ 0x416 b7ad0: 0x13fc10e _ZN19File_OpenDedupeFile21handleExtentListWri

2 Needless error messages similar to the following in the server log for Deduplication enabled filesystems being replicated. 2011-04-22 13:40:27: DEDUPE: 3: PE: PersistState() returned error: Filesystem is a read-only 2011-04-22 13:40:27: DEDUPE: 3: PE: WriteOperationBegin filesystem fsid 292 is read only. There were no dedupe operations initiated by the Celerra administrator at the DR site.

Code has been fixed to hide these informational messages

413576 / 39901452

FileMover

The following entries describe fixed FileMover related problems:


Tracking Number

1 Incremental backup sometimes picked up unchanged FileMover (DHSM) files

Corrected the way offline file size is calculated so it will not cause incremental backup to incorrectly backup unchanged data.

405361/ 37027684

File systems

The following entries describe fixed issues pertaining to File Systems.


204

Version 6.0.42.3


1 The /nas/sbin/uc_config -verify command returned the incorrect translation settings for the client specified. This could lead to thinking the configuration is not configured correctly.

/nas/sbin/uc_config -verify now always gives the correct translation setting being used for the specified client.

387225 / 36947724

1 While using Networker 6.1 to backup the file system via a CIFS connection, the FLR log was flooded with a large amount of repeating log entries. The rapidly expanded FLR log contributed to extending backup times. Following is an example of the repeating FLR entries: <date> Extend Retention Period on worm committed file : Inode No = <Inode> : Generation No = <generation> : UID = <UID> : Gid = <GID> : File Mode = <mode> : Filesize = <size> : Old RP = <date> : New RP = <date> : Failed

Code no longer logs a message when the retention period of a locked file in FLR-C file system matches the retention period that is passed in by requests.

406428 / 39472638

1 An NFSv4 state object was recycled prematurely (one reference missing) which caused a Data Mover to panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x000068cfbe Err code: 0 Target addr: 0x00000001c0 **.The stack trace showed: 6760: 0xbf96c0 _ZN7DskDump9checkdumpEPKc+0x7a 67a0: 0xbf98ce dump_return+0x4 67b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 67d0: 0x14ce5d fault_dump+0x89 67f0: 0x13378d handle_fault+0x1d f7f9d0: 0x6a3257 _Z17NFSv4_lockExpiredP11NFSv4_State+0x15 f7f9f0: 0x6a32c2 _ZN17NFSv4_delegClient9sendBreakEjiyy+0x 30 f7fa50: 0x874603 _ZN10Delegation9sendBreakEjyy+0x5f3

The Data Mover now confirms the release of the object before allowing it to be used.

410524 / 40054076

1 A race condition occurred when some of the file system operations resulting in vnode closure occurred while the same file system was being unmounted on the Data Mover,

Code has been fixed to properly release all the references on the

356875 / 33841178

205

Version 6.0.42.3


resulting in a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../ufs.hxx at line: 1077 : UFS hashRemove: not on hash chain. The stack trace showed: 0xd1cedbb0: 0x13d890 waitForReboot+0x90 0xd1cedbd0: 0x13da8d fault_dump+0x67 0xd1cedbf0: 0x13d98d PANIC+0x29 0xd1cedc00: 0x61cd54 _ZN9UFS_Vnode4loseEv+0x8e 0xd1cedc20: 0x5f05f1 _ZN9VFS_Vnode7reclaimEv+0x81 0xd1cedc40: 0x5ee99e _ZN9VFS_Vnode11getNewVnodeEv+0x40 0xd1cedc50: 0x5ee337 _ZN9VFS_VnodenwEmP14VFS_FileSystem+0x9 0xd1cedc70: 0xb98871 _ZN15DHSM_FileSystem8newVnodeEm8dirTypes +0x41

file system prior to the unmount, preventing this condition.

While unmounting a file system with a DFS root share on it, a race condition occurred which led to a Data Mover panic with the following DART panic/fault message: >>PANIC: I/O not progressing LastVol touched not Known. The stack trace showed: 0x802fe90: 0x14124c waitForReboot+0x90 0x802feb0: 0x1414d5 fault_dump+0x67 0x802fed0: 0x1413bd PANIC+0x29 0x802fee0: 0x8b3c56 _ZN9RawIO_IRP13checkProgressEv+0x11a 0x802ffd0: 0x8b3b32 _ZN17IrpProgressThread5startEv+0x6 0x802ffe0: 0x145bfa Sthread_startThread_internal+0xe

This race condition has been eliminated by locking the share only when we are sure it is unmounting file system.

383368 / 38343174

206

Version 6.0.42.3


2 Memory was consumed and not freed when a wildcard search (ex. dir *.txt) was run from an FTP session with no matches. This led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 289 : Out of memory. The stack trace showed: 0xe10d1c30: 0x14124c waitForReboot+0x90 0xe10d1c50: 0x1414d5 fault_dump+0x67 0xe10d1c70: 0x1413bd PANIC+0x29 0xe10d1c80: 0x13bef6 _ZN19Memory_AddressSpace10allocPagesEi12 Memory 0xe10d1ca0: 0x1459f1 _ZN7SthreadC1EPcP10DQueueHeadIS_Eji+0x1e b 0xe10d1cc0: 0xfc4f42 _ZN21ckptBlockAllocatorThr6launchEPKc+0x 6a 0xe10d1cf0: 0xf9f333 _ZN19PoolManagedDeltaVol14startAllocator Ev+0xa 0xe10d1d10: 0xf981a5 _ZN19PoolManagedDeltaVolC1EPKcS1_P7LVolu meS3_m

The memory was not correctly freed after each wildcard search within FTP. This has been fixed.

407209 / 39528390

1 An NFS client issued two illegal cross file system rename requests to the Data Mover at the same time which led to a deadlock condition and loss of NFS access.

Code was modified to reject the illegal request and return the NFS3ERR_XDEV error when it receives cross file system rename request.

402586 / 38891086

1 Linux NFSv4 clients exposed to the issue described in https://bugzilla.redhat.com/show_bug.cgi?id=264721 present the same clientid string to the Data Mover and therefore can be seen as a single client. This caused a race condition which led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 1ce857 Err code: 0 Target addr: c8 **. The stack trace showed:

The fix is a protective measure which takes an exclusive mutex before revoking the locks held by a client. This will avoid the Data Mover panic, but users should upgrade the Linux clients’ kernel to

400183

207

Version 6.0.42.3


0x2c62c84: 0x141278 waitForReboot+0x90 0x2c62ca4: 0x141501 fault_dump+0x67 0x2c62cc4: 0x13a398 handle_fatal_fault+0xa 0x2c62d00: 0x67ca57 _ZN11NFSv4_State5closeEmP8stateid4P8nfs4 _ctxRP4 0xe1059d20: 0x67b0d8 _ZN16NFSv4_StateTable5CloseEmP8stateid4P 8nfs4_ 0xe1059d50: 0x669c7d _Z10nfs4_closeRP4msgbS1_RPhR8nfs4_ctx+0x 18b 0xe1059d90: 0x665e30 _Z23nfs4_convertNfsv3Statusm+0x626 0xe1059ee0: 0x647bdd _ZN17nfs_serverInitial5startEv+0xad7

the versions mentioned in the Bugzilla report to resolve the issue completely.

2 On write heavy workloads, a lot of the following messages "Run out of staging buffers" were observed indicating that the current static setting of 6 for agregateWork::numbuf was not sufficient during peak write loads.

Introduced a new parameter that can be tuned to get rid of 'Run out of staging buffers' messages. This newly introduced parameter ufs.aggregateC ommitNbufs can be adjusted to handle peak write loads for iSCSI and uncached IO. The parameter values are as follows: 32 (default), 6 (minimum) and 256 (maximum). Refer to the Parameters Guide for more detailed information about this parameter.

409307

208

Version 6.0.42.3


1 After a file system corruption, FSCK discovered the corruption and fixed most of it by freeing the corrupted inodes. However, fsck could not properly handle some special file types (block type & char type) from the file system perspective. Further access to such file types caused the Data Mover to panic with the following DART panic/fault message: isAValidIndBlkBuf: bad bn. The stack trace showed: 0xdc2bb740: 0x14124c waitForReboot+0x90 0xdc2bb760: 0x1414d5 fault_dump+0x67 0xdc2bb780: 0x1413bd PANIC+0x29 0xdc2bb790: 0x778cb5 _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0xdc2bb950: 0x78f1aa _ZN14UFS_FileSystem17isAValidIndBlkBufEP 13Buff 0xdc2bb9d0: 0x785233 _ZN9UFS_Vnode17readIndirectBlockElRP13Bu ffer_C 0xdc2bba00: 0x7877ae _ZN9UFS_Vnode8bmapListEmmPmPi+0x228 0xdc2bba70: 0x78a5e6 _ZN9UFS_Vnode15blocksReadAsyncEjiPP13Buf fer_Ca

FSCK code has been enhanced to properly handle block and char type of the files and prevent this panic.

404536 / 39140368

1 When file system had many corrupted directory entries and FSCK was running on this file system, the Data Mover panicked with the following DART panic/fault message: *** GP exception. Virt ADDRESS: 87db94. Err code: 0 ***. The stack trace showed: 0x2c3be94: 0x14124c waitForReboot+0x90 0x2c3beb4: 0x1414d5 fault_dump+0x67 0x2c3bed4: 0x13a398 handle_fatal_fault+0xa 0x2c3bff0: 0xe8155d31 _end+0xe495d509 0xe8151fc0: 0x87d471 _ZN12Notfoundlist6locateEmP6dirValRbPPS_ +0x3f 0xe8152040: 0x87d471

The FSCK code has been changed to handle large number of corrupted directory entries correctly.

391164 / 37582944

209

Version 6.0.42.3


_ZN12Notfoundlist6locateEmP6dirValRbPPS_ +0x3f 0xe81520c0: 0x87d471 _ZN12Notfoundlist6locateEmP6dirValRbPPS_ +0x3f 0xe8152140: 0x87d471 _ZN12Notfoundlist6locateEmP6dirValRbPPS_ +0x3f

1 An out-of-sequence redundant close operation for the same owner led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../include/stream.hxx at line: 989 : datab::freeRef: db_ref <= 0. The stack trace showed: 0xe00dbc30: 0x14124c waitForReboot+0x90 0xe00dbc50: 0x1414d5 fault_dump+0x67 0xe00dbc70: 0x1413bd PANIC+0x29 0xe00dbc80: 0x1be7cc _ZN5datab7freeRefEv+0x32 0xe00dbc90: 0x674ee3 _ZN11NFSv4_Owner9saveReplyERP4msgb8nfsst at4P8n 0xe00dbcc0: 0x679376 _ZN11NFSv4_State5closeEmP8stateid4P8nfs4 _ctxRP 0xe00dbd20: 0x6778fe _ZN16NFSv4_StateTable5CloseEmP8stateid4P 8nfs4_ 0xe00dbd50: 0x6665e5 _Z10nfs4_closeRP4msgbS1_RPhR8nfs4_ctx+0x 18b

Corrected the serialization of close operations to prevent this Data Mover panic.

401063 / 38621014

2 Under heavy load, the quota usage value was incorrectly reported as being different between on-disk and in-memory. The error message in the server_log implied that the on-disk value was corrupted. 2011-01-24 18:11:47: 13157531648: UFS: 3:[SNXE-FS-02131] inode usage record is not consistent: on disk 2915057 in memory 2915058 type 1 id 32773 ouId 0 2011-01-24 19:28:15: 13157531648: UFS: 3:[SNXE-FS-02131]

This error message will be printed only if there is an actual discrepancy between the on-disk and in-memory quota values.

402598 / 38854944

210

Version 6.0.42.3


1 During a FSCK of a file system, the Data Mover panicked if a block address contained the bad block flag rather than a real address. DART panic/fault message: >>PANIC in file: ../DeltaIO.cxx at line: 727 : DeltaIO_Read_CB::failed. The stack trace showed: 0xd939f430: 0x14124c waitForReboot+0x90 0xd939f450: 0x1414d5 fault_dump+0x67 0xd939f470: 0x1413bd PANIC+0x29 0xd939f480: 0xf9a361 _ZN15DeltaIO_Read_CB6failedE14BlockIO_St atus+0 0xd939f490: 0x8b3d33 _ZN9RawIO_IRP9irpFailedE14BlockIO_Status y+0x21 0xd939f4b0: 0x8c8253 _ZN11HyperVolume17readBlockInternalEP10D iskIO_ 0xd939f510: 0xf99d69 _ZN15DeltaIO_Read_CB10doneCommonEv+0x239 0xd939f570: 0xf99b1e _ZN15DeltaIO_Read_CB14getMappingDoneEy+0 x82

FSCK will now skip the read when the block address has been flagged as bad.

393729 / 37862428

1 A Data Mover panicked when a file system was mounted with the options: ntcredential and accesspolicy=NT or SECURE. DART panic/fault message: *** GP exception. Virt ADDRESS: 6d4599. Err code: 0 ***. The stack trace showed: 0x2c67d28: 0x141278 waitForReboot+0x90 0x2c67d48: 0x141501 fault_dump+0x67 0x2c67d68: 0x13a398 handle_fatal_fault+0xa 0x2c67e20: 0x6d440d _ZN15File_NamingNode8accessOKEmP17Access _Creden 0x34509e10: 0x64c1cd _ZN16worldAccess_List16printWorldAccessE P9secF 0x34509ee0: 0x647afd _ZN17nfs_serverInitial5startEv+0xad7 0x34509fe0: 0x145c26 Sthread_startThread_internal+0x

Code was performing an extra release on NT credential for UNIX in the case of file system mounted with ntcredential and accesspolicy=NT or SECURE. This has been fixed.

407971

211

Version 6.0.42.3


e

1 When running FSCK on a corrupted file system, FSCK may cause memory corruption and a Data Mover may panic when FSCK tries to repair certain types of corruption. DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 351 : Memory corrupted. The stack trace showed: 0x23a81dc0: 0x14124c waitForReboot+0x90 0x23a81de0: 0x1414d5 fault_dump+0x67 0x23a81e00: 0x1413bd PANIC+0x29 0x23a81e10: 0x1fe8e7 free+0x69 0x23a81e40: 0x1fefb8 _ZdlPv+0x10 0x23a81e50: 0x1fefce _ZdaPv+0xc 0x23a81e60: 0x24880e _ZN13cMemoryBufferD1Ev+0x1e 0x23a81e70: 0x24a2fb _ZN11cMemBufListD1Ev+0x15

The fix adds additional protection in FSCK code to avoid memory corruption.

399957 / 37582944

2 Server_2 hung and did not respond to commands, client did not have access to data. A failure to release all TCP streams as NFSv4 connections were being closed eventually resulted in there being no TCP streams available on the Data Mover to service commands or user I/O requests.

The fix ensures all TCP streams are released when NFS v4 clients disconnect.

412602 / 40332336

1 For Redhat Linux, the nfs.v3xfersize transfer size was limited to 32K.

In order to have better NFS performance from Linux RedHat client running on VMware, the maximum nfs.v3xfersize is now tunable by using the param nfs v3xfersize=131072 command.

373951

212

Version 6.0.42.3


1 After a client rebooted, a Data Mover panicked with the following DART panic/fault message: Assertion failure: 'refLock.isExclusiveLocked()' on line 1691 in file '../nfsv4_state.cxx' >>PANIC in file: ../assert.c at line: 27 : assert failed. The stack trace showed: 0x11456ba70: 0x2a7d0b _assert+0x63 _ZN15NFSv4_ClientEnt7ConfirmEyPA8_KhR8nf s4_ct 0x11456bb10: 0x7dd53d _Z19nfs4_create_sessionRP4msgbS1_RPhR8nf s4_ct 0x11456bc70: 0x792ec5 _Z9nfs4_linkRP4msgbS1_RPhR8nfs4_ctx+0x16 87 0x11456be40: 0x81fea2 _ZN17nfs_serverInitial5startEv+0x1e40 0x11456bfd0: 0x16ad51 Sthread_startThread_internal+0x7

NFSv4.1create proper client mutex and prevents this condition.

403427

1 An empty tree quota directory was removed by an unexport operation. This occurred if a name was specified when the export was created.

When processing an unexport, code now performs the remove operation only when the last component is a symbolic link and not a real directory.

388560 / 37329964

1 Under heavy NFSv4 load, NFSv4 clients may log "Remote I/O error" messages and responses from the Data Mover may be interrupted for 5-10 minutes. The network trace would show NFS4ERR_RESOURCE errors from the Data Mover. The error occurs because the code did not release NFSv4 lock owners correctly.

The fix allows early revocation of NFSv4 lock owners to free up resources, and avoid hang conditions.

402884

1 When parameter mount.forceFullShowmount was set to 0, Data Mover MAC threads became blocked. GUI and CLI commands became unresponsive, and the following message was logged in the server_log periodically: ThreadsServicesSupervisor: Service:MAC Pool:MACExec BLOCKED for xxx seconds:

Corrected a deadlock that occurred when this param was set to zero and 1) server_export command was run, and 2)

403999 / 39088798

213

Version 6.0.42.3


Server operations may be impacted.

another process attempted to generate a list of FS mounts (such as the server_mount command) simultaneously.

1 User set large security descriptors (SD) with redundant ACEs. When code tried to read those SDs after several days, it ran into a memory panic because the SD needed a large chunk of memory. Although there was enough memory, code could not grant memory allocation requests because of memory fragmentation. This caused the Data Mover to panic with the following DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory. The stack trace showed: 40bc10: 0x145966 _ZN19Memory_AddressSpace10allocPagesEl12 Memory 40bc50: 0x24a34e malloc+0x11a 40bc90: 0x24ae1b _Znwm+0x1b 40bcb0: 0x24ae46 _Znam+0x4 40bcc0: 0xcf9e07 _Z11querySDInfoP13smb_threadCtxR11smb_Ou tMngt 40bdd0: 0xdd00e7 _ZN14smb2_threadCtx16replyQuerySDInfoEP4 msgb+ 40be70: 0xdd0bd0 _ZN14smb2_threadCtx14replyQueryInfoEP4ms gb+0x 40bea0: 0xdcbc33 _ZN14smb2_threadCtx7processEP4msgbi+0x68d

Added a threshold param cifs.acl.sdcompactthreshold. If a SD is larger than this parameter's value, code tries to compact it by removing redundant ACEs. The default is 16KB. This is to mitigate allocation problem in a fragmented memory.

404188 / 38935580

2 Code was accepting non $DATA stream types, which are not compatible with NTFS. When trying to copy *.jpg files from a Celerra CIFS share to a local Window XP client, the following error was displayed: cannot copy xxx: The filename, directory name, or volume label syntax is incorrect.

Code no longer accepts non-$DATA type ADS.

155357 / 31498524

214

Version 6.0.42.3


Install/Config/ Upgrades

The following entries describe fixed issues related to installations, configurations and upgrades.


Tracking Number

2 Upgrade process might modify the value of some NDMP parameters as the maximum value of these parameters have been changed in newer releases. A new PUHC check is required to display a warning message regarding this modification.

A new check has been included to warn user if any NDMP value would be deprecated during upgrade.

394516

2 If a failure occurs during installation, or the system is rebooted at the beginning of the install process, the NBS service does not start properly at the next reboot.

With this fix, the /etc/nbs.conf file is updated at the same time as the connection setup. Any reboots after that, will include a successful nbs service restart.

401432 / 38569244

2 PAHCPUHC might fail with a hard drive failure error even when the hard drive is not in faulted condition. ------------Errors-------------------- Control Station: Check the hard drive in the control station Error HC_CS_14505083128: The 'smartctl -H' command failed. Action : Re-run "/usr/sbin/smartctl -H /dev/hda" to observe the output. Contact your system/network administrator to resolve the issue or escalate this issue through your support organization. ------------------------------------------

The check was relying on the output of the smartctl command instead of the return value.

Code changes have been made to avoid the PAHC/PUHC failure when the hard drive in the Control Station is not in faulted condition.

381917

215

Version 6.0.42.3



Tracking Number

1 Upgrade is seen to fail in the "Wait for NAS service to start" task. This occurred when a lot of access commands, to the file systems mounted on the Data Mover, were being issued from the Control Station while the upgrade was in progress.

The fix is to stop audited for the duration of the upgrade, and restart the same with the pre-upgrade settings at the end of the upgrade.

388554

NDMP

The following entries describe fixed Network Data Management Protocol/Backup problems:


Tracking Number

1 VBB file level restore of offline/stub files did not work. When calling setAttr() for those files, VBB set file size which caused the file to be recalled and hence become online.

Code was modified to set attributes without changing file size.

395232 / 37208224

2 Some NDMP backup automatic checkpoints were left behind and had to be removed manually. This occurred if the Celerra was very busy and the checkpoint creation completed after the timeout expired.

The checkpoints created automatically by an NDMP backup are now removed

377919 / 35499712

Networking

The following entries describe fixed issues related to Networking.


Tracking Number

216

Version 6.0.42.3


1 During a Data Mover reboot, it was possible that the SLIC did not power up and caused the Data Mover to fail bootup. This could lead to a data unavailability situation.

This was fixed by upgrading the firmware on the SLIC.

406453

1 Under rare circumstances, the Data Mover may not reply to all NULL requests from a client. This occurred if the transmit queue for the TCP session was full.

The Data Mover will now reply to every NULL request, even if the transmit queue to that client is full.

404522 / 38052726

2 When having many simultaneous FTP connections to a Data Mover, some of the connections failed due to a small listen backlog queue. When this occurred, Data Mover logged the following error in the server log: Too many connections in port 21's listen queue.

Code has been changed to remove the limitation of small listen queue. It is now unlimited and can be controlled with the param tcp.backlog.

405484 / 38895674

2 When using IPv6, the host cache table can monotonically grow as a result of expired host entries. These expired entries can make user route retrieval times exceed the watchdog timer and cause a Data Mover panic with the following DART panic/fault message:>>PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG. The stack trace showed: eee0: 0x14dd29 system_watchdog+0x61 eef0: 0x132913 t_nmi+0xc3 efd0: 0x155263 Sthread_startThread_internal+0x7

The fix adds support to remove stale IPv6 hosts from routing and neighbor caches and prevent this panic.

405168 / 39158304

NFS

The following entries describe fixed issues pertaining to Network File Systems.

217

Version 6.0.42.3



Tracking Number

1 When a file system was close to 100% full, there was incorrect logic in the space allocation code which caused the Data Mover to allocate more space on the file system for a new writes than was available. This led to a Data Mover panic with DART panic/fault message: alloc failed: counters out of sync. The stack trace showed: 0xe1321840: 0x13fb94 waitForReboot+0x90 0xe1321860: 0x13fda9 fault_dump+0x67 0xe1321880: 0x13fc91 PANIC+0x29 0xe1321890: 0x7286c7 _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0xe1321a10: 0x737932 _ZN9UFS_Vnode5allocEP20UFS_UpdateDescrip torllj 0xe1321a50: 0x735cfc _ZN9UFS_Vnode10ballocListEP20UFS_UpdateD escrip 0xe1321af0: 0x73979a _ZN9UFS_Vnode15blockChainWriteEiP10Block Chaini 0xe1321bc0: 0x73910d _ZN9UFS_Vnode15blockChainWriteEiP10Block Chain+

The allocation logic code has been fixed to ensure this panic no longer occurs when a file system is close to full.

159227

218

Version 6.0.42.3



Tracking Number

1 While deleting a large file, an overflow in a file system data structure caused a memory corruption issue which resulted in a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000957f75 Err code: 2 Target addr: 0x1000000008 ** The stack trace showed: a760:0xbfb6b0 _ZN7DskDump9checkdumpEPKc+0x7a a7a0:0xbfb8be dump_return+0x4 a7b0:0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot a7d0:0x14ce5d fault_dump+0x89 a7f0:0x13378d handle_fault+0x1d 05610:0x886740 _ZN14UFS_FileSystem13getCgBlkEntryEj+0x10 05620:0x8a003a _ZN14UFS_FileSystem9hashallocEP20UFS_Upd ateDes 05690:0x8a0335 _ZN14UFS_FileSystem8allocBlkEP20UFS_UpdateDesc

Code has been fixed to properly handle this condition and eliminate the panic.

402066

1 When using NFS V4, a bug in the Linux clients made multiple clients look as a single client to the Data Mover. A race condition in this context caused the Data Mover to panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 1ce857 Err code: 0 Target addr: c8 **. The stack trace showed: 0x2c62c84: 0x141278 waitForReboot+0x90 0x2c62ca4: 0x141501 fault_dump+0x67 0x2c62cc4: 0x13a398 handle_fatal_fault+0xa 0x2c62d00: 0x67ca57 _ZN11NFSv4_State5closeEmP8stateid4P8nfs4 _ctxRP4 0xe1059d20: 0x67b0d8 _ZN16NFSv4_StateTable5CloseEmP8


411712

219

Version 6.0.42.3



Tracking Number

stateid4P 8nfs4_ 0xe1059d50: 0x669c7d _Z10nfs4_closeRP4msgbS1_RPhR8nfs4_ctx+0x 18b 0xe1059d90: 0x665e30 _Z23nfs4_convertNfsv3Statusm+0x626 0xe1059ee0: 0x647bdd _ZN17nfs_serverInitial5startEv+0xad7

1 NFS lock was not preserved upon a Data Mover failover.

Code now preserves the statd registry when file systems are unmounted by administrator.

407035

Replication

The following replication-related bugs that were fixed in this release:


Tracking Number

2 If the Control Station timed out while waiting on a Replication Stop operation and retried, the operation hung and blocked further operations on the session.

When the Data Mover receives a Stop request, it now checks whether the Replication Sessions is already in cleaning-up state and if so returns a TaskInProgress error.

404635

220

Version 6.0.42.3



Tracking Number

1 When one checkpoint with the same snapVol ID as PFS Vol ID was inactivated, the PFS version was wrongly inactivated. User was unable to refresh or create any checkpoints on their main production file system. The following error was generated: Error: DpRequest_vs_InternalError.

The product version is now skipped when version for inactivated checkpoint is searched in inactivateVersion.

404437 / 39049562

1 Data Mover panicked with DART panic/fault message: PANIC in file: ../DeltaVol.cxx at line: 7585 : Value in TOC is different from the value in btree. After the failover, some checkpoints were inactive.

A flaw in index slot relocation process that can cause an inconsistency in checkpoint metadata has been fixed.

414584 / 40596048

2 An attempt to delete a Replication V2 session hung if the command was issued while a failover operation was in progress.

The delete operation will not be accepted while the Replication session is in a clean-up state and the operation will fail with Abort in progress error to the user.

402024

1 Under heavy load, in configurations with remote iSCSI Replication with many iSCSI LUNs and snaps, Replication Manager job may timeout due to the Data Mover not responding in time to iSCSI snap create request from Replication Manager.

A new mechanism was added to Data Mover code and Replication Manager code in order to handle slower response times under heavy load. Now Replication Manager can

396493

221

Version 6.0.42.3



Tracking Number

query the status on Data Mover and if Data Mover cannot response fast enough, RM will retry the operation until it succeeds or until configurable time out value is reached. This fix is available in Replication Manager version 5.3.3 or later.

1 When a SnapSure SavVol becomes full, the oldest checkpoint needs to be unmounted and inactivated in order to free space. If a refresh of the oldest checkpoint occurs during this process it can result a deadlock, blocked threads, or a DART panic with the fault message PANIC: I/O not progressing.

Code now waits for the completion of the umount before taking a reference on the checkpoint file system.

399556

2 While using Replicator V2, if the ACK from the Data Mover failed/delayed to reach the Control Station, the Control Station retried sending the modify request. There was a problem in the code which caused a memory leak and cmd reference leak. Hence, there was an inflight task running on the Data Mover.

Code now returns TASK_IN_PROGRESS directly to the Control Station to prevent the memory and reference leak.

403651

Security

The following security-related bugs that were fixed in this release:

222

Version 6.0.41.4



Tracking Number

1 Netgroups that are nested and begin with a special character were not being parsed.

Nested netgroups that begin with special characters are now fully supported.

402867

2 After creating the LDAP or storage domain user, the account type was displaying as "local" instead of LDAP in the GUI.

This issue occurred because the resource file contained a white space followed by the value. White spaces are now removed from the resource file for all account type values.

406555

Storage

The following entries describe fixed storage-related issues:


Tracking Number

1 When a NAS was installed at location where SAN hosts had in-use FAST pools, Solutions Enabler (SE) code incorrectly parsed some float numbers as integer in the FAST pool data area. This caused SE discover failure and NAS install failure if there were existing SAN hosts with FAST pools. Error IM_14505279983 : Failed to synchronize with new backend storage

This issue was fixed in SE code by changing from integer parsing to float parsing for several values.

400965 / 38569244

Version 6.0.41.4

Network File System

The following entries describe fixed issues related to Network File Systems:

223

Version 6.0.41.4



Tracking Number

1 When a checkpoint of the replication destination file system became corrupted and user was accessing the destination file system while replication data transfer was in progress, the read requests were redirected to the corrupted checkpoint. This caused a Data Mover panic with the following DART panic/fault message: D2320489_0: DeltaIO_Read_CB failed due to blk 16 corruption. The stack trace showed: 0x21a53f50: 0x14124c waitForReboot+0x90 0x21a53f70: 0x1414d5 fault_dump+0x67 0x21a53f90: 0x1413bd PANIC+0x29 0x21a53fa0: 0xfc2c9b _ZN22CkptCorruptMarkRequest13handleReque stEv+0 0x21a53fc0: 0xfc2b26 _ZN23CkptCorruptMarkerThread5startEv+0x80 0x21a53fe0: 0x145bfa Sthread_startThread_internal+0xe

Code no longer redirects read requests to a corrupted checkpoint.

402304

Replication

The following entries describe fixed Replication issues:


Tracking Number

224

Version 6.0.41.4



Tracking Number

1 If certain operations were performed on multiple Replication V2 sessions at the same time, the destination file system became out of sync with the source if a transfer was in-progress when the operation was initiated. The operations that can cause this issue are failover, stop, start -reverse and delete (if the session wasn't previously stopped). When one of these operations caused a transfer to be interrupted, the destination file system was restored to the state before the transfer is started. Only one of these restores could run at a time so while one was still in progress others failed and logged a message in the sys_log similar to: Apr 1 15:00:00 2011:DART:REP:ERROR:8:Slot 2:::1301670000:Destination=100_CK200064400 000_0000_200_CKM00100200000_0000(alias=fs1 ), restoring a checkpoint failed: DpRequest_vs_VersionExist.

The code has been fixed. Multiple file systems can now be restored to their state before the current transfer is started. This allows operations to be safely run in parallel for multiple Replication V2 sessions.

406391 / 39554202

1 In some cases during bootup, the code might incorrectly decide that not enough space was reserved in order to recover a checkpoint. This can cause the Data Mover to panic and the checkpoints for the file systems to be inactive. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: fde91e Err code: 0 Target addr: 58 **. The stack trace showed: 0x2c62684: 0x141278 waitForReboot+0x90 0x2c626a4: 0x141501 fault_dump+0x67 0x2c626c4: 0x13a398 handle_fatal_fault+0xa 0x2c62700: 0xfbedcd _ZN19PoolManagedDeltaVol13startFlushLogE P10Delt 0x200816d0: 0xfbbabc _ZN19PoolManagedDeltaVol9indexSyncE18ind exSync 0x200817b0: 0xfd36ca _ZN10SnappedVol14notifyChildrenEN7LVolum e8noti

Code has been fixed to ensure checkpoints are recovered correctly.

399188

225

Version 6.0.41.3



Tracking Number

0x200817e0: 0x8cee38 _ZN11HyperVolume14notifyChildrenEN7LVolu me8not 0x20081800: 0x8c7687

Version 6.0.41.3

CEE

The following fixed issues are related to Celerra AntiVirus Agent, Celerra Event Enabler, and Celerra Event Publishing Agent:

Severity

Symptom Description

Fix Summary

Tracking Number

1 When using Celerra Event Enabler, CIFS users were not able to access CIFS shares during multiple FS scans.

Created a new parameter, viruschk.fsscanburst, which defines the number of files sent in one burst by the scan process to the Virus Checker engines. The default is 20 files. The file system scanning will be suspended until the number of files reaches the value set. If the parameter is set to 0, the burst will be dynamically set to the number of AV thread times the number of online CAVA servers.

386093

226

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 When stopping a virus checking service on a Data Mover while having a lot of unchecked files in virus checker queue, the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 846 SYSTEM WATCHDOG. The stack trace showed: 0xe6475ed4: 0x14124c waitForReboot+0x90 0xe6475ef4: 0x1414d5 fault_dump+0x67 0xe6475f14: 0x1413bd PANIC+0x29 0xe6475f24: 0x1432c9 system_watchdog+0x77 0xe6475f34: 0x139f6a _t_nmi+0x68 0xe6475f70: 0x71dcc6 _ZN15File_NamingNode14setCheckStatusEi+0 x34 0xe6475fb0: 0xd10d60 _ZN13check_initial5startEv+0x19a 0xe6475fe0: 0x145bfa Sthread_startThread_internal+0xe

Code has been fixed to properly handle the locks within the code and prevent this panic.

388970 / 37306154

2 When virus checking was enabled on a Data Mover, it caused failover to take additional time to unmount the Data Mover's root file system. This added approximately 1 minute to failover/ shutdown time.

Virus checking no longer takes a reference on the root file system, so its unmount is not delayed during shutdown/failover.

387885 / 37272266

1 The Data Mover was unable to send viruschecking requests to an AV server using MSRPC when a GPO update was trying to connect to a slow or non-responsive domain controller.

New CIFS client connections are no longer blocked by other unrelated connection attempts in progress.

390569 / 37333686

1 When using the nonotify mount option on a file system for which virus checker was configured, a deadlock condition occurred

Code has been fixed and no longer tries to scan a file if it cannot

395401 / 38038272

227

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

if a client tried to delete a file in the same directory where any other file was waiting to be scanned.

get the path using the triggerset.

2 Following a manual check of a file using Symantec virus checking engine, the ctime of the file was changed. This caused a differential backup done by NDMP to be larger than expected.

Code has been fixed to properly update the access time without changing the ctime, so it no longer affects differential NDMP backups.

400155 / 36478750

1 When a file type (.gz) was added to the exclusion list of the virus checker config file viruscheck.conf, the Data Mover still continued to scan these files. This issue occurred when .gz files were already in the pending queue to be scanned, and the filter was changed while the virus checker service was stopped. Once the service was restarted the Data Mover incorrectly sent the files in backlog queue to be scanned.

The Data Mover now removes files from the pending queue when this type of files is added to the exclusion list and virus checker is restarted.

392790


The following are fixed problems related to the Celerra Management Applications, i.e., Celerra Monitor, Celerra Manager, WebUI, or Celerra MMC Snap-ins:

Severity

Symptom Description

Fix Summary

Tracking Number

2 User could not use Celerra Manager to change checkpoint schedule from a monthly to weekly recurrence. The following error was displayed: Modify schedule <schedule name>. The task scheduler is not responding. Please try again later.

Code was fixed to allow the modification of a monthly schedule to weekly schedule.

378575 / 36073712

228

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 If a Data Mover was renamed (ex. server_2 was renamed to datamove2), the nas_alerterd could not monitor its CPU utilization.

Fixed monitoring of Data Movers that are renamed.

404632

1 A Celerra system can only be managed by a version of Unisphere that has a Celerra plug-in of the same version or later. If Unisphere was downloaded from one of the Storage Processors in the array, an older version of the Celerra plug-in was installed and the 6.0.40-5 Celerra was not be manageable. A message of Unsupported or Newer Version was displayed.

Code has been fixed to allow all in-family management through Unisphere.

395531

2 Rapid rebooting or restarting a Data Mover caused a process leak in the Control Station.

Implemented a fix to properly close child processes to prevent process handle leaks,

398424

2 When user queries for the capacity info or the capability info of the file systems, if there were checkpoints, XML API displayed the following error message and not the correct value: Unexpected exception of unknown nature.

This issue occurred because VPFS file system did not have storageMB value and a null check was not performed. If the file system is of type - VPFS, no details of the file system will be displayed.

384126

2 During an enumerate instance for an association, the SMI-S response received contained trailing spaces. There was no code to trim the APL returned value for serial.

Added code to remove the trailing spaces from the serial value.

385415

2 The device type 10 gigabit and the speed 10000FD were not added in the code. When APL returned this device type and speed, XML API displayed the value as other and incorrectly sent the files in backlog queue to be scanned.

Added the new device type and the new value of speed to the code.

386449

229

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

3 The last boot time on the Data Mover properties page was not localized whereas the current time was localized.

Dates are no longer localized on the Data Mover properties page.

381447

2 User could not see and confirm that an NTP server was configured. When CSA queried about ntpServers attached to a Data Mover, it did not list any, even though one was attached.

The APL query attribute for fetching NTP server info contained an error which has been fixed.

387693

2 User received the following error message when using the Celerra CIFS management snapin on the Data Mover: Internal Error: Unable to retrieve GPO setting from server. Please insure that the cifs server is still connected to the domain. Possible solution would be to rejoin the domain.

This issue was caused by a typo in an XML response name value which resulted in the setting not being correctly interpreted by the Celerra Management MMC. This has been fixed.

394296

2 User was unable to add Domain Mapped Groups that contained spaces to Unisphere. When trying to setup LDAP with Unisphere to map domain groups, if user tried to add a group with spaces, an error was displayed which states only _ . - are valid special characters.

Spaces in domain mapped users / groups are now supported in Unisphere.

396496 / 37607682

2 When a CIFS share spanned across multiple CIFS servers, XML API identified only one server and sent the request to APL. APL expected all the servers as the key, and returned the following error: Archiver::createDhsmConnections() Failed to query the file system of '<name>' for server '10.10.17.153'. Error Reason = 'XmlApi::recvXmlResponse Error.

The code has been modified to check for share name and CIFS server name and return the information correctly.

137720

230

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 On the Checkpoint Extend page, after selecting the Extend from Meta Volume option, the following message was displayed: No volumes are available' even though there were free meta volumes available. This occurred because meta volumes with the same disk type as the file system were listed in the dropdown on the Checkpoint Extend page rather than same disk type as checkpoint.

The GUI code has been modified so that filtered meta volumes have same disk type as the checkpoint and not the file system.

369551

2 Warning message was not generated and visible on the dashboard of the Unisphere manager or sys_log if FTP configuration was incorrect in /nas/site/automaticcollection.cfg file.

Added additional logging in the code path to generate the warning correctly if it cannot contact the remote FTP server.

403294

2 When the migration file system related migration encountered an error, for example, the migration source NFS server was down, its state was displayed as 'Idle' instead of 'Failed'.

The MGFS state is now reported correctly.

376681

2 User created an nfs export with options using the CLI, for example, server_export server_3 –o root=root_hosts,rw=root_hosts:fr-hosts,access=root_hosts:fr-host s. When viewing the export through Unisphere, it was displayed as root:hosts instead of root_hosts.

Fixed the APL code that caused underscores to be replaced with colons.

399488 / 38452644

2 When the number of Celerras being managed was more than 5, the Celerra Capacity View Panel in the Unisphere dashboard showed incorrect information in the title and incorrect display order of the Celerras in the chart.

Changed the order in which the Celerras capacity details objects passed before updating the view.

372223

2 Celerra/CUT plugin reported a different set of domain members upon startup than the CLARiiON plugin.

CimomAccess.pm treated the IPs in reverse order as a new entry and added it to domain_list file. The

374045

231

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

CimomAccess.pm file was modified to make sure it identifies IPs in reverse order as well and does not populate the domain_list file.

2 When using Unisphere and clicking on a list page button (e.g. Delete, Properties, etc.) user waited for the expected Confirm Delete Dialog page to pop up. Occasionally, the Confirm dialog popped up for a split second then disappeared behind the Unisphere window. On a slow or busy client, this happened more often.

Unisphere code was fixed so that the action pop-up window remains in the foreground above the main window.

378218

2 The creation of user quotas from the root of a file system could not be performed from the GUI, but worked fine when using the CLI. The option "none" was missing from the user and group creation pages in Unisphere.

Code was modified to include the "none" option.

379661

2 The Run CLI command in Unisphere was not localized. There were no leading/trailing character strings for the text and none of the buttons were converted.

This has been fixed, and the Run CLI command is now localized.

381117

1 The checkpoint delete popup in Unisphere came up blank when running with a language pack enabled.

Code has been modified to check for the localized version of the string.

394269

2 User could not modify the Maximum Storage Capacity through Unisphere, but could modify it using the CLI. The following error was returned: invalid size specified.

This issue occurred because APL did not send the unit as Megabytes (MB). This has been fixed.

382336

232

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 In Celerra Manager Group Quota page users were unable to search quotas by GID or group name.

Search by GID and group names capability was added to Group Quota page of Celerra Manager.

379910 / 36289838

CIFS

The following entries describe fixed issues related to Common Internet File Systems:

Severity

Symptom Description

Fix Summary

Tracking Number

1 The Data Mover panicked and with a Page Fault Interrupt when a pre-released version (BETA) of MACOS X 10.7 (Lion) client accessed a CIFS share. The stack trace showed: 0x2ab3a08: 0x13fb44 waitForReboot+0x90 0x2ab3a28: 0x13fd59 fault_dump+0x67 0x2ab3a48: 0x138d68 handle_fatal_fault+0xa 0x2ab3a70: 0xb22430 _ZN21cMsRPC_LsarpcSimulate13lsarProcess2 DEP13sm 0xc3745b60: 0xb29164 _ZN21msrpc_legacyOperationI21cMsRPC_Lsar pcSimu 0xc3745b90: 0xc9ffd1 _ZN22msrpc_interfaceHandler13callOperati onEtR1 0xc3745bf0: 0xc9edb5 _ZN15msrpc_serverCnx15_processRequestER2 3msrpc 0xc3745cc0: 0xc9df72 _ZN15msrpc_serverCnx10processMsgEP13smb_ thread

The code was modified to correctly handle the LSA Lookup call 0x2D and avoid this Page Fault condition

406316/ 39423800

1 When accessing NAS files from Windows, the count of hard links was incorrect. When using the perl stat command on a Windows file, the link count on the file was always 1.

Code now returns the correct link count for the file (note: for directories, the link count is always 1).

392782

233

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 If a new file filter file was created with Windows XP or Windows 2003, it worked as intended for users with the same systems, but not on Windows 7 or Windows 2008 . For these newer Windows versions, files could be created but could not be deleted nor edited.

Code has been fixed to deny creation of files based on file filtering rules for Windows7/ Windows 2008 clients.

395700 / 37640396

1 During a domain SID migration, an invalid SID in a Security Descriptor (SD) caused a NULL referencing condition. This led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: bf1d26 Err code: 0 Target addr: 2 **. The stack trace showed: 0x2acfc64: 0x13fb94 waitForReboot+0x90 0x2acfc84: 0x13fda9 fault_dump+0x67 0x2acfca4: 0x138d94 handle_fatal_fault+0xa 0x2acfdb0: 0x7a9656 _ZN11AclDatabase5parseEP23ParseAclDataba seConte 0x56b7ff60: 0x7a8eed _ZN11AclDatabase5parseEP23ParseAclDataba seCont 0x56b7ffb0: 0x7a84c1 _ZN23ParseAclDatabaseContext5startEv+0x6 9 0x56b7ffe0: 0x144344 Sthread_startThread_internal+0xe

Code now ignores the invalid SDs during a domain SID migration and the migration continues.

391200

234

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 When using DHSM CIFS connection, recall of some files failed with the following: MGFS: 3: error direct read block 7 of entry 4248, status 20 – EndOfFile. The issue occurred because a secondary CIFS server returned a non-zero size for getAttr request, but a zero size for Open request.

Code now handles cases where Open and getAttr request sizes do not match.

384678

2 Copying a file over a DFS link failed from an SMB2 client, reporting there was not enough space, even though there was plenty of space on the target of the DFS link. This is because the size of the DFS root file system was being checked rather than the target of the DFS link.

The Data Mover now sets the REPARSE_POINT file attribute in the DFS link node. This attribute directs the SMB2 client to query for file system size of the DFS link target.

401201 / 38573658

1 Unable to access server when Microsoft security policy in raised Windows 2008 R2 domain. See Microsoft article: 2345886, Description of the update that implements Extended Protection for Authentication in the Server service, http://support.microsoft.com/default.aspx?scid=kb;en-US;234588 6

Use correct CIFS-based SPN to access Domain Controller.

393548 / 37839090

2 When migrating SIDs from one domain to another using server_cifs –Migrate command, unknown SIDs were not migrated.

The sever_cifs -Migrate command has been fixed, and is now will include unknown SIDs.

387781 / 37493702

1 When setting permissions from a Windows client, the special permission "Apply onto" as "Files only" did not get propagated to child directories.

Permission inheritance is now handled correctly for "Files Only".

383935 / 36713302

1 User experienced slow performance with Windows7 client using MSOffice applications when Celerra CIFS server was configured with UNIX security. Access to shares enumeration returned

Added support for share information through SRVSVC pipe, granted access to unprotected functions,

389804 / 36737358

235

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

an internal error status, causing delays on the client side.

like shares enumeration.

2 When using CIFS auditing and only enabling events for Read operations, Data Mover generated logs for ALL events, regardless of configured events. The issue occurred because of an error in the mechanism that checks if an audit event log must be generated.

The error has been corrected, and the Data Mover will now generate logs only for specified events.

384108 / 36692600

1 Under certain stress circumstances, the Domain Controller replied to LSARPC requests with an error: STATUS_INSUFFICIENT_RESOURCES. The Data Mover kept attempting to use this DC, never getting any information. Server log contained errors like: LsarOpenPolicy2::Request failed, status=INSUFFICIENT_RESOURCES. From the client standpoint, SID was displayed instead of clear text names when attempting to get ACL content.

The connection with the overloaded Domain Controller will be dropped. The Data Mover will now reconnect to another Domain Controller.

382103 / 36576304

1 User experienced experience slow performance with Windows7 client using MS Office applications when Celerra CIFS server was configured with UNIX security.

Added support for share information through SRVSVC pipe to accommodate pipe paths with ASCII names.

383876 / 36737358

2 When SMB2 was enabled on Data Mover, widelinks stopped working, and all links appeared as files in Windows Explorer.

The code has been fixed to allow widelinks when SMB2 is enabled.

381269

1 When using CEPP/CAVA over MSRPC and a CEPP or CAVA server did not respond to an SMB thread, other SMB threads were locked awaiting the timeout of the CEPP/CAVA client. This appeared as hangs or slow access for users.

The code now allows the single SMB thread timeout without holding a lock on all other SMB threads.

379264 / 36219350

236

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 Attempting to copy with EA (NTFS extended attribute) from a Window machine to Celerra using SMB2 (Win7/W2K8-R2 client and SMB2 enabled on Celerra) fails with popup message: The extended attributes are inconsistent or The file or directory is corrupt or unreadable.

Celerra does not support EA but the status being returned on SMB2 was not appropriate (STATUS_NO_EAS_ON_FILE). Celerra now reports STATUS_EAS_NOT_SUPPORTED. Also, the parameter used to make Celerra ignore EAs and not return an error (cifs set_eas_ok) now applies to SMB2.

380397 / 36177168

1 When deleting a CIFS server ,the timeout of the Control Station command must be set to at least 120 seconds. Instead, it was set to 50 seconds. In some cases 50 seconds was insufficient for DART to perform the deletion. As the result, the Control Station timed out and failed to delete a server from netd and boot.cfg. Upon Data Mover reboot, the deleted CIFS server was unexpectedly recreated on the Data Mover.

The timeout value for delete CIFS server request to the Data Mover has been correctly set to 180 seconds.

374845 / 36219350

2 If the server_thread command was used with -all, the Data Mover panicked with the following DART panic/fault message: *** GP exception. Virt ADDRESS: 0x00006f3ed1. Err code: 0 ***. The stack trace showed: 0x3685760: 0xbf93d0 _ZN7DskDump9checkdumpEPKc+0x7a 0x36857a0: 0xbf95de dump_return+0x4 0x36857b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 0x36857d0: 0x14ce5d fault_dump+0x89 0x36857f0: 0x13378d handle_fault+0x1d

NFS threads now correctly display the last operation executed.

390668

237

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

0x106585940: 0x15a5bf _ZNK15cThreadActivity33displayIdleThread Info_ 0x1065859f0: 0x6f3e8e _ZNK17nfs_serverInitial21displayIdleThre adInf 0x106585a00: 0x15af4a _ZN15cThreadActivity11parseThreadER12cSv cPars

1 User could not access a server from an

NT4 machine (or any client like Samba using old NTLM SessionSetup request). This occurred when NTLMV2 was enforced with: LAN Manager Auth Level: Send NTLMv2 response only, refuse LM and NTLM NTLM SSP Minimum Client Security: Require NTLMv2 session security NTLM SSP Minimum Server Security: Require NTLMv2 session security NT4 client setup: LAN Manager Auth Level on NT 4 client configure to “Level 2 – Send NTLM authentication only”. The following entries were observed in server_log: 2010-07-30 10:22:55: 13157007360: SMB: 3: Client=customerClient NTLMV2 authn required 2010-07-30 10:22:55: 26041909248: SMB: 6: Client=customerClient killed by NTLMV2 required 2010-07-30 10:22:55: 26041909248: SMB: 6: SessSetupX failed with RPC_NT_UNSUPPORTED_AUTHN_LEVEL

Code now allows authentication requests to be sent to the domain controller even if it is suspected to be NTLMV1, based on size.

386340 / 3620668

238

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A Data Mover panicked or the viruschecker stop operation hung when viruschecker was restarted multiple times. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000155833 Err code: 2 Target addr: 0x0000000008 **. The stack trace showed: 0x3686760: 0xbf90b0 _ZN7DskDump9checkdumpEPKc+0x7a 0x36867a0: 0xbf92be dump_return+0x4 0x36867b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 0x36867d0: 0x14ce5d fault_dump+0x89 0x36867f0: 0x13378d handle_fault+0x1d 0x1af6be20: 0x2069bd _ZN8RWlockRT10sharedLockEj+0x65 0x1af6be60: 0xf72de4 _ZN12virusChecker5mountEP14VFS_FileSyste m+0x20 0x1af6bf00: 0xf6bb96 _ZN12check_survey5startEv+0x416

Code now waits until the end of the viruschecker survey thread before restarting viruschecker. Also removed an extra shared lock that led to this deadlock condition.

388629

2 With Access Based Enumeration (ABE) enabled, some DFS links may not appear in directories even though the user has the right to see them.

The problem is now fixed and a user can see all DFS links if he has the right to see them.

392030 / 37302982

2 The offline function on a CIFS share did not work as expected for Vista and Windows 7 clients. When the offline setting was disabled on a CIFS share, a Vista or Windows 7 (SMB2) client still could use offline function on this folder.

The offline flags of the CIFS share are now properly returned to the SMB2 clients.

386604 / 36952502

1 An out-of-memory panic occurred when using large credentials (containing more

Code now allocates less contiguous

388145 /. 37290766

239

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

than 1500 groups from the forest) for NTLM authentication.DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 289 : Out of memory. The stack trace showed: 0xf84bd310: 0x13fb94 waitForReboot+0x90 0xf84bd330: 0x13fda9 fault_dump+0x67 0xf84bd350: 0x13fc91 PANIC+0x29 0xf84bd360: 0x13a8f2 _ZN19Memory_AddressSpace10allocPagesEi12 Memory 0xf84bd380: 0x1f6c3a _Z11more_memoryjP9free_list+0x4a 0xf84bd3a0: 0x1f6cf8 malloc+0x98 0xf84bd3d0: 0x1f7485 _Znwm+0x15 0xf84bd3e0: 0x1f74b2 _Znam+0xc

memory to complete user's NTLM authentication.

240

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A memory corruption in a Kerberos context was used by a function that then returned a NULL pointer. The pointer was not checked before being passed into another function, which then tried to dereference the pointer, resulting in a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 1fa53e Err code: 0 Target addr: 0 **. The stack trace showed: 0x2acb158: 0x13fb94 waitForReboot+0x90 0x2acb178: 0x13fda9 fault_dump+0x67 0x2acb198: 0x138d94 handle_fatal_fault+0xa 0x2acb1c0: 0x56e267 krb5_cc_resolve+0x17 0x1eca92b0: 0x56e609 krb5_cc_default_suffix+0xa3 0x1eca92e0: 0x53e1db _Z26krb5_gss_release_principalPjPPv+0x46 b 0x1eca93a0: 0x53ed90 _Z25krb5_gss_acquire_cred_extPjPv22gss_b uffer_ 0x1eca9400: 0x5456ff gss_acquire_cred_ext+0x4b

The fix is to check the return value of the pointer, and return an error if the value is NULL.

383732

3 When the .etc of a share's file system was missing, each time the share was mounted for the first time, a message similar to the following was logged to the server_log: Cannot get .etc dir (28=InternalError).

This message was informational only, as the code was correctly handing the issue. This message is now hidden from the user.

393943 / 37873442

1 When trying to set ACL to a file which had an empty DACL (Discretionary Access Control List) using emcsetsd, the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: /disks/NSG_Root/Maint_4/5.6_build_nfs_env/ Dart/

Before the fix, code was overwriting the header of the ACL. Code now checks if the file has an empty DACL, and write the new ACL after the ACL header.

397384 / 38188396

241

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

server/src/cfs/include/acl.hxx at line: 1522 bad SD storage size. The stack trace showed: 0xde7617e0: 0x141254 waitForReboot+0x90 0xde761800: 0x1414dd fault_dump+0x67 0xde761820: 0x1413c5 PANIC+0x29 0xde761830: 0xe824cc _ZN14set_celerra_sd7processEv+0x336 0xde761910: 0xe826e9 _Z21cifs_set_celerra_sd_1P27cifs_set_cel erra_s 0xde761990: 0xe85ab9 _ZN29xattrp_IterSvc_Parent_InitialC1EP16 Rpc_Sv 0xde761a40: 0x4a9092 _ZN21Rpc_Svc_Child_Initial13svc_getreqse tEP7rp 0xde761c30: 0x4ac16f _ZN25Rpc_IterSvc_Child_Initial7svc_runEv +0x10b

242

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 User could not install CelerraCifsMgmt on a Windows 64-bit OS.

Removed the restriction in the CelerraCifsMgmt install package that prevented user from installing it on 64-bit system. The following message is now displayed when installing CelerraCifsMgmt on a 64-bit system: You are now installing 32-bit CelerraCifsMgmt on a 64-bit Operation System. Celerra UNIX property page extensions are only available in Active Directory Users and Computers on 64-bit Operation System through the following instructions: 1. From the Start menu, select "Run ..." 2. In Open field, input the command: mmc /32 dsa.msc 3. Click the OK button to launch the Active Directory Users and Computers

374292

2 A user application might not work properly when using temporary files on a CIFS share. The Celerra returned a SHARING_VIOLATION instead of DELETE_PENDING.

The fix respects the precedence of the DELETE_PENDING on the SHARING_VIOLATION status.

391399 / 37403078

2 During data migration from Windows 2003 file servers to CIFS servers on Celerra using EMCOPY, the following error appeared: EMCOPY: permission changes either copies or does not copy files from

Added /secforce option that force emcopy to overwrite destination security and prevent this condition.

393429 / 37414066

243

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

source to dest.

2 Using the nonotify mount option did not prevent notifications of changes to a directory from being sent to CIFS clients. Also, the option could interfere with directory renames.

The nonotify mount option now disables CIFS notifications and does not interfere with directory renames.

395058 / 37393392

Control Station

The following entries describe fixed problems related to the Celerra Control Station.

Severity

Symptom Description

Fix Summary

Tracking Number

1 When an NBS session (which included failovers) was started on the Control Station, a race condition occurred in the NBS driver that led to a Control Station kernel panic and reboot with the following message: Kernel panic - not syncing: Fatal exception.

To fix this issue, a locking mechanism was added to prevent this race condition.

144684 / 29744606

2 The server_ldap -lookup command did not query groups containing spaces in the name.

The Control Station now correctly processes LDAP queries containing spaces in the group name.

374688 / 35603200

3 When a nas_quotas report command was entered with incorrect parameters, the command returned the following error message instead of the command usage: nas_quotas: Warning 17716809847: server_2: NIS Error: UID= <file system>.

With the fix the command now rejects incorrect parameters and displays command usage.

376711 / 35626866

244

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 If the interface name of the server_cifs command ended with CRLF which is "carriage return" on Windows , the Eiffel trace and vdm.cfg or netd table was corrupt.

Code now truncates the CRLF character before doing transaction.

380335 / 36355420

2 Management daemon crashed while attempting to allocate memory in order to service an XML message, causing the core dump to fill up the /nbsnas partition.

In a rare case where the memory is fragmented to the point that the service of an XML message was unable to complete, a retry mechanism was added to re-attempt to allocate the memory. Memory is now recovered in the background.

386399 / 36932966

2 User was denied permission error if the username exceeded 8 characters while executing the new CLI which needed special admin role privilege.

Code has been fixed to verify the username and password correctly.

377427 / 35919230

1 When a VG8 system was configured with 3 Data Movers, a false alert that the power supply was removed from non-existent Data Mover was reported. 0x330070051 Error: Local emc70pri common DME 1 Power Supply B0: Removed.

Code now correctly detects the components present in the enclosure.

397608 /; 38148792

2 If user specified a storage system for a file system extension and the storage system was different than what the file system already exists on, they might get storage from the wrong storage system.

If a storage system is specified in file system extension, code now only takes storage from that storage system for the extension.

378214

3 When creating or refreshing a checkpoint task from APL, if the scheduler failed, both APL task manager and APL scheduler generated the same error: Create checkpoint ckpt_alert_sched_001

When creating/refreshing checkpoint task from APL scheduler fails, only one error will be

378439

245

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

of schedule alert_sched has failed.

displayed.

3 When running a dbchk -f -q command, the following error was logged in the sys_log: A Celerra database error was detected. An inconsistency in the Celerra database files was detected. Run the $NAS_DB/tools/dbchk tool to identify the exact error. However, when user ran the dbchk tool, the output was clear: [nasadmin@apoh0010cs03 ~]$ /nas/tools/dbchk -pvwx dbchk: done.

If the nas command finishes within 10 seconds, dbchk will no longer report an error (unless nasdb finds some inconsistency problems).

395926 / 38096334

1 The Data Mover reported too many errors in server log for software correctable errors and generated SMI (system management interrupt). The system clock got out of sync while the SMI was being processed, which resulted in a Data Mover panic with the following DART/panic fault message: >>PANIC in file: ../sched.cxx at line: 1541 : SYSTEM WATCHDOG. The stack trace showed:

0x3635a40: 0x14c81d _ZN12Memory_Frame13setProtectionEm19Memo ry_Protection_t+0x23d 0x3635a60: 0x14ce29 fault_dump+0x55 0x3635a80: 0x14cf02 PANIC+0x5a 0x3635aa0: 0x14dd29 system_watchdog+0x61

0x3635ab0: 0x132913 t_nmi+0xc3 0x3635b68: 0x378674 _ZN12bcmnicDevice12bcmInterruptEPv+0x5a 0x3635b98: 0x1d338f _ZN3ISR6invokeEv+0x2d 0x3635bb8: 0x1d3e91 shared_isr+0x207

Fixes the PCI correctable error handler for devices on the motherboard so it correctly prints warning messages instead of panicking the system and falsely indicting the CPU module.

374413

246

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 Under heavy system load, a large amount of data frames were dumped, causing dumpsize to exceed 24G and dump_slot not showing the right size. This could result in the dump being corrupted, or important data could be missed from the dump.

Code has been fixed to skip the buffer data frames when not required and calculate the correct data dump size.

389936

2 When each checkpoint schedule was fired off, APL scheduler checked its next execution time. If the next_execution_time + 60 seconds < current time, this schedule was marked as miss, and the checkpoint did not create/refresh. If too many schedules fired off at the same time, some of them did not create/refresh the corresponding checkpoint.

Increased the timeout value from 60 seconds to 5 minutes to allow more schedules to run at once.

384138 / 35719928

2 The server_ifconfig server_X -a command did not display netname for IPV6 interfaces. Netnames were displayed for IPv4 interfaces only.

Corrected the query sent to the Data Mover ensure it displays both IPv6 and IPv4 netnames.

384477

2 User changed the physical Data Mover's name to the same name as the VDM using the server_name command. This caused several server_xxx commands to not function properly.

Code now rejects an attempt to change the physical Data Mover's name to the same name as the VDM's.

385998 / 37053750

2 Using the nas_dbtable with a filter on a field which was not the primary key, or intensive use of the nas_dbtable caused the Data Mover to run out of memory

This memory leak condition has been eliminated.

394952

CORE

The following entries describe fixed problems related to Celerra Data Movers.

247

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 Commands against the Data Mover failed with Error 5: server_4 : Input/output errorTimed out or the Data Mover did not respond to the command at all because the MAC threads were hung in an unmount deadlock. [root@A00DRFL003 log]# server_date ALL server_2 : Tue Dec 1 10:04:11 GMT+1 2009 server_3 : Tue Dec 1 10:03:43 GMT+1 2009 server_4 : Error 5: server_4 : Input/output errorTimed out server_5 : Tue Dec 1 10:04:41 GMT+1 2009

ReclaimFreeNode and giveUp are now called directly without checking for a useCount of zero.

16063 / 32384702

248

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

A memory leak in the rpcbind_getaddr() function of the code caused a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../bufpool.cxx at line: 389 : couldn't get a free page. The stack trace showed:

0xcd2ab8f0: 0x13f244 waitForReboot+0x90 0xcd2ab910: 0x13f441 fault_dump+0x67 0xcd2ab930: 0x13f341 PANIC+0x29 0xcd2ab940: 0x17d7ef _ZN18Buffer_Cached_Pool17makeCachedBuffe rsEi+0 0xcd2ab970: 0x13b4ed _ZN19Memory_AddressSpace16makeCachedBuff erEv+0 0xcd2ab990: 0x17d37d _ZN18Buffer_Cached_Pool15getAnyCachedBuf Eji+0x 0xcd2ab9c0: 0x6b1ec3 _ZN14UFS_FileSystem8getBlockEliRPv+0x41 0xcd2aba00: 0x6b1629 _ZN14UFS_FileSystem15readBlocksAsyncElii PP13Bu

The code has been changed to avoid the memory leak.

380233

CSA

The following entries describe fixed issues pertaining to the Celerra Start-up Assistant.

Severity

Symptom Description

Fix Summary

Tracking Number

2 While configuring the Control Station through CSA, if user tried to set the CS and blade time to (GMT +5:30) Asia/Calcutta (as shown in the drop-down menu). The "Apply Changes" step in CSA failed with NTP server, time zone and unicode properties set up on primary

Time zone information has been changed in the CSA to reflect the updated spelling of Kolkata.

379078 / 36174238

249

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

blade... - Invalid time zoneccmdEscColon Asia/Calcutta.

1 The site ID that was entered by CSA incorrectly used double quotes and was not visible in Celerra Manager or Unisphere.

Code was fixed to ensure users can see the information in Celerra Manager or Unisphere.

361995

2 CLARiiON events for NX Systems with AX4-5F8 backends were passed to /var/log/messages but were not passed to sys_log as expected.

Code was altered to send these CLARiiON events to the sys_log.

393090

1 When the user tried to upgrade a Fibre Channel (FC) enabled systems from version 5.6 to 6.0 or tried to install integrated systems (running NAS and FLARE), the Celerra Startup Assistant failed to perform the Celerra and CLARiiON system domain join operation.

The Celerra Startup Assistant was modified to perform a domain join for FC-enabled systems.

382973

2 In the Celerra Startup Assistant (CSA), during Control Station install, the Gateway field was not indicated as a required field.

The Gateway field is now marked as a required field.

361899

Deduplication


Severity

Symptom Description

Fix Summary

Tracking Number

250

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 Error messages similar to the following were incorrectly logged as ERROR in the server_log when mounting the FLR file system for the first time: Error 2010STORAGENumpaths option u, E EDT 2010FSTOOLSFLR data popula, Error T 2010FSTOOLSflr_clock = -1, flr_max_time.

These errors were informational only, and are no longer displayed.

378549

1 During a read operation of a file that has recently failed deduplication through the CIFS compression or DHSM API methods (via VMWare plugin), the Data Mover panicked with the following DART panic/fault message: isAValidIndBlkBuf: bad bn.The stack trace showed:

0x271b1450: 0x141254 waitForReboot+0x90 0x271b1470: 0x1414dd fault_dump+0x67 0x271b1490: 0x1413c5 PANIC+0x29 0x271b14a0: 0x77663b _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0x271b1660: 0x78ccbe _ZN14UFS_FileSystem17isAValidIndBlkBufEP 13Buff 0x271b16e0: 0x782d47 _ZN9UFS_Vnode17readIndirectBlockElRP13Bu ffer_C 0x271b1710: 0x7852c2 _ZN9UFS_Vnode8bmapListEmmPmPi+0x228

Code was updated to ensure data is properly flushed to disk before deduplication is started via the CIFS compression method or the DHSM API method (via VMWare plugin) to prevent this panic.

378009 / 36040202

1 There is a deadlock situation in the F-RDE re-ingest code that may be hit if a Data Mover panics and reboots or a file system is re-mounted while having entries in the re-ingest queue. The deadlock will prevent files in the re-ingest queue from being processed. This may cause the re-ingest queue to become filled and reject new entries.

Code has been fixed to prevent this deadlock condition.

379193

251

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 When using NDMP backup on file systems that have deduplication enabled, NDMP backups could take more space than necessary, and NDMP jobs could fail with the following error messages in the log DEDUPE: 3: EDRS: RP CSX COMPRESSOR: internal error - decompression failed: 2 DEDUPE: 3: EDRS: TRANSFORMING FILEDATA UNIXFSIS: readBlob: transformOut failed at offset 3244032: RDE_STATUS_DECOMPRESSION_ERROR. Because of a limitation in the compressor module of code a failure could occur if this code receives more than 2 requests for decompression at a time.

Code now ensures that compressor module code does not receive more than 2 requests for decompression at a time.

381922 / 36502216

1 When using Deduplication, if many CIFS clients were performing directory scanning on F-RDE enabled file systems, CPU utilization on Data Mover reached 100% and impacted performance. The issue occurred because code in deduplication policy engine (PE) was not optimized.

PE code has been optimized and will not cause 100% CPU utilization under these conditions.

385163 / 36929424

2 There was a race condition between removal of deduplicated files and file system unmount operations. Under certain conditions, deduplicated data space was not reclaimed from the file system when files were removed. A message was displayed in the server log similar to: 2010-05-09 09:40:57: 13163036672: DEDUPE: 3: Cannot find file system with id 15, cannot cleanup object <dart://rde ...

If you attempt an unmount during removal of deduplicated files, the code now waits for threads to complete before performing the unmount.

367398

252

Version 6.0.41.3


File system

The following entries describe fixed file system issues:

Severity

Symptom Description

Fix Summary

Tracking Number

1 When a file system was mounted on a VDM, if the tree root’s inode number was the same as the inode number of the mount point to which that VDM was mounted, renaming such tree quota from the client resulted in “/..” path reported.

Code has been fixed to properly handle this condition so the rename of quota tree will be completed as expected.

387997

1 The cylinder group bitmap initialization process is started asynchronously for each file system as they are mounted on the Data Mover. If a file system was unmounted or the Data Mover was rebooted, failed over, or failed back before this operation completed, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 7831ec Err code: 0 Target addr: 2c **. The stack trace showed:

0x2ba5e34: 0x140cec waitForReboot+0x90 0x2ba5e54: 0x140f71 fault_dump+0x67 0x2ba5e74: 0x139eec handle_fatal_fault+0xa 0x2ba5ee0: 0x1c35ab _ZN17SvcThread_InitialIS_I15Sthread_Init ialEE5s 0xe6ca1fe0: 0x1455d4 Sthread_startThread_internal+0xe

The cgbitmap initialization is now aborted if the file system state is unmounting, unmounted, frozen, or corrupt. It will be reinitialized when the file system is remounted.

357994

1 Data Mover denied write I/O with NFS4ERR_ACCESS instead of NFS4ERR_LOCKED in when a file is locked.

NFS threads now return the correct error code.

392703 / 37513228

1 When a file was being truncated through setAttr or setLength calls and any panic or unexpected reboot happened during the

Code has been fixed to properly handle SetAttr/setLenght logic

382208

253

Version 6.0.41.3


same time, the Data Mover panicked when attempting to access that file. DART panic/fault message: isAValidIndBlkBuf: bad bn Panicking Processor: 3. The stack trace showed:

0x79b313f0: 0xa6bb4c _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0x79b31560: 0xa927dc _ZN14UFS_FileSystem17isAValidIndBlkBufEP 13Buff 0x79b315e0: 0xa83009 _ZN9UFS_Vnode17readIndirectBlockEiRP13Bu ffer_C 0x79b31630: 0xa8407d _ZN9UFS_Vnode4bmapEiRiPiS1_+0x1c1 0x79b316a0: 0x18c38a1 _ZN10DHSM_Vnode14itruncInternalEP20UFS_U pdate 0x79b31710: 0xa7fbd1 _ZN9UFS_Vnode6itruncEP20UFS_UpdateDescri ptoryi 0x79b317a0: 0xa80208 _ZN9UFS_Vnode8inactiveEPvi11File_Statusi i+0x47 0x79b31820: 0x1893362 _ZN10MGFS_Vnode8inactiveEPvi11File_Statu sii+0

to prevent the panic under this condition.

1 A Secure NFS mount from a client could caused a Data Mover panic with the following panic header: Virt ADDRESS: 0x00006dd730 Err code: 0 Target addr: 0x000000000c.

Code was updated to prevent the panic and ensure Secure NFS mounts succeed.

398127

254

Version 6.0.41.3


1 During file system operations, code improperly passed a handle to a different function. Later the handle has changed, but the function that this handle was previously passed to did not realize this, and used the old handle which led to memory corruption. This resulted in the panic with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 525 : Memory corrupted. The stack trace showed:

4056c0: 0x24aa1f free+0x79 405700: 0x24aded _ZdlPv+0x9 405710: 0x9682b9 _ZN20UFS_UpdateDescriptor8completeE11Fil e_Sta 405820: 0x8a1c9f _ZN21aggregateCommitExtent6doWorkEP9VFS_Vn ode 405870: 0x930b47 _ZN13aggregateWork9writeListEP10gatherList P9V 4058f0: 0x92f6d4 _ZN17VersionFile_Vnode15commitAllocListEP1 1ex 405940: 0x8aafa5 _ZN9UFS_Vnode18uncachedWriteChunkEyjRP4msg bRi 405b30: 0x8ab627 _ZN9UFS_Vnode13uncachedWriteEyjP4msgbiRj+0 x1a

Code now properly passes the handle by reference to other functions.

397129

1 After a Data Mover failover/failback, certain client requests (allocating writes and file creates) were blocked until all cylinder group (CG) bitmaps were initialized. Depending on the configuration, background initialization took a few minutes or panicked after 6 minutes with the following DART panic message: >>PANIC in file: ../sched.cxx at line: 1543 : SYSTEM WATCHDOG. The stack trace showed:

Panicking Processor: 0 0x730cd070: 0x14ee4b system_watchdog+0x61 0x730cd080: 0x140efb t_nmi+0xc3 0x730cd1b0: 0x8d929b _ZN14UFS_FileSystem8allocInoEP20UFS_Upda teDesc 0x730cd300: 0x9436e3 _ZN9UFS_Vnode12getChildNodeEP2

The block allocate function was modified to only wait until any one cylinder group is found with free blocks prior to continuing the client requests.

399405

255

Version 6.0.41.3


0UFS_Updat eDescr 0x730cd3a0: 0x8e615b _ZN9UFS_Vnode9setLengthEyiiiP13Buffer_Ca ched+0 0x730cd410: 0x150ceb6 _ZN10DHSM_Vnode9setLengthEyiiiP13Buffer_ Cache 0x730cd460: 0x7ee121

2 In the xlt.cfg file, when the subnet address

was defined in the format “x:y:z:w, a:b:c:d”, it failed to lookup the host address and ignored the subnet entry.

Code was fixed by adding delimiter support for "," along with the existing "/".

375781

2 When FSCK was run on a file system that had bad sectors, a memory leak occurred which caused the Data Mover to reboot.

The allocated bad inode objects and bad sector objects are now deleted.

114569

1 User got Out-of-Space errors on virtually provisioned file systems that started off at a small size (1 GB) and had Auto Extension enabled during a full clone operation.

Full clone operations are now suspended during auto extend. If file system Auto Extension is enabled and the HWM is exceeded, code pauses the block copy loop until the extension completes (usually around 30 seconds). In the event we really do run out of pool space and can't complete the extend we only allow 5 delay tries during the full clone operation.

383532

1 A resume event was not processed when all asynchronous input-output threads were busy deleting large sparse files. The file system became inaccessible during mount cycles and the user saw the following error: No such file or directory>>File system inaccessible.

The file system reference is no longer stored until the resume event completes. If there is a request to unmount the file system, it can now be processed immediately.

378345

256

Version 6.0.41.3


1 When creating new tree quotas and copying large amounts of data, if there were not enough sequential pages for quotaTreeVnodes array, the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 486 : Out of memory. The stack trace showed:

562ee0: 0x145966 _ZN19Memory_AddressSpace10allocPagesEl12Me mor 562f20: 0x24a3de malloc+0x11a 562f60: 0x24aeab _Znwm+0x1b 562f80: 0x24aed6 _Znam+0x4 562f90: 0x8caddb _ZN14UFS_FileSystem20treeQuotaOn_Interna lEPcS 563210: 0x8cf8b7 _ZN14UFS_FileSystem16quotaOn_InternalEPcS0 _+0 563300: 0x833066 _ZN14VFS_FileSystem7quotaOnE9quotaTypetPcS 1_+ 563350: 0x8293d0 _Z17File_QuotaControlR6OptionP18VFS_FileSy stem

Code was optimized so that sequential pages are no longer required for this operation.

399030 / 38339946

1 During a Data Mover failover in NAS 5.6, the internal mount command had become case sensitive. File systems with accesspolicy=mixed (lower case - upgraded from older NAS code) failed to mount and printed the following message in the cmd_log.err - 2009-10-31 15:26:38.777 db:0:31381:E: /nas/sbin/slot_recover 2 -panic 2009-10-31 15:30:54.196 db:0:31381:E: /nas/sbin/slot_recover 2 -panic: replace_storage: replace_mount: eng0:mount: unknown accesspolicy: mixed. home0:mount: unknown accesspolicy: mixed. corp0:mount: unknown accesspolicy: mixed. derived0:mount: unknown accesspolicy: mixed. This failure caused the system to abort the

Reinstated case insensitive behavior for the server mount accesspolicy options during failover.

158137 / 32026672

257

Version 6.0.41.3


failover and restore all services on the original Data Mover.

1 The Data Mover panicked because of the UFS Log recovery mechanism attempted to invoke FSCK resulting in NULL pointer dereference. DART panic/fault message: *** Invalid Opcode exception. Virt ADDRESS: c ***. The stack trace showed:

0x2c297e8: 0x141254 waitForReboot+0x90

0x2c29808: 0x1414dd fault_dump+0x67 0x2c29828: 0x13a3a0 handle_fatal_fault+0xa 0x2c29880: 0x8504b1 _ZN7Fsck_SB4readEP11NewFsckArgp+0x51 0x507b9e0: 0x84f8f7 _Z8check_sbP11NewFsckArgpP7Fsck_SB+0x23 0x507ba70: 0x85c139 _Z9doNewFsckP9IO_ObjectiiiiRPci+0xed 0x507bac0: 0x85d8b8 _Z20kickOffFsckInRecoverP15fsRecoverStru ct+0x84 0x507baf0: 0x81a850 _Z20handleRecoverCommandR6Option+0x60a

The code now checks for a valid file system object for FSCK kickoff during ufs log recovery.

376227 / 35807576

2 User was unable to delete rep v2 iSCSI replication session when its production file system was full. The task hung and the following error was displayed in the server_log: VCS: 3: Failed to write attribute file, fsid=31, path=fs31_T1_LUN2_CK200082600835_0000 name=.opaque status=NoSpace

Code was modified to ensure the session delete completes under this condition.

359609 / 34143216

1 The Data Mover incorrectly returned NFS3ERR_NOENT (No such file or directory) when the checkpoint (.ckpt) directory was being looked up.

The fix eliminates the race by no longer caching checkpoint mountpoint names.

162103 / 32681254

258

Version 6.0.41.3


1 A Data Mover panicked when moving a directory to new parent which was being deleted. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 6d6db8 Err code: 0 Target addr: 44 **. The stack trace showed:

0x2c2a524: 0x141254 waitForReboot+0x90 0x2c2a544: 0x1414dd fault_dump+0x67 0x2c2a564: 0x13a3a0 handle_fatal_fault+0xa 0x2c2ab30: 0x81cfe9 _ZN19ShadowDirectoryBase6renameEPtmPS_S0 _P17Acc 0xdd52bcf0: 0xc515d0 _ZN8AppLibNT6renameEPtS0_lR11File_Status R10ext 0xdd52bed0: 0xc427c2 _ZN14smb1_threadCtx11replyRenameEP4msgb+ 0x11e 0xdd52bf40: 0xbe1033 _ZN14smb1_threadCtx7processEP4msgbi+0xcdf 0xdd52bfa0: 0xbd5820 _ZN13smb_threadCtx5startEv+0x6d0

If a parent directory to which the directory is being moved has been deleted, code now returns "no such file" and no longer panics.

371168 / 35185054

1 User reported poor write response time with VMware ESX using Celerra through NFS with uncached file systems.

In this case, the Data Mover serialized the writes and sent the replies after the last write storage. Code was changed to no longer serialize incoming NFS write requests.

404209

1 When writes were performed on a large (>2TB) file system which was nearly full (>80%), threads become blocked. This was observed by slow user access and messages like the following in the server log: ThreadsServicesSupervisor: Service:NFSD Pool:NFSD_Exec BLOCKED for 419 seconds. A panic is also possible with the message: PANIC in file: ../ufslog.cxx at line: 6895 : sync: no

The cylinder group bitmap which is used to quickly find free blocks is now initialized at bootup.

386406 / 37135080

259

Version 6.0.41.3


progress reducing dirty list.

1 Using IP addresses with an invalid format (separated by a dot instead of a comma, for example, 102.30.121.131.102.30.109.8) as part of the export command caused memory corruption. The Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 7cd814 Err code: 0 Target addr: 1d3 **. The stack trace showed:

0x2c29dd4: 0x141254 waitForReboot+0x90 0x2c29df4: 0x1414dd fault_dump+0x67 0x2c29e14: 0x13a3a0 handle_fatal_fault+0xa 0x2c29e70: 0x7cd63f _ZN14UFS_FileSystem17startMetadataSyncEv+0 xf 0xb278ff40: 0x7cd4dc _ZN14UFS_FileSystem4syncEi+0x260 0xb278ff90: 0x7cc395 _ZN21UFS_SyncThreadInitial5startEv+0x2c7 0xb278ffe0: 0x145bd6 Sthread_startThread_internal+0xe

Code now limits the buffer overflow that can potentially happen with an incorrect IP address format.

370824 / 35147654

1 A user file system reported corruption of the gidmap file which prevented it from being mounted on the Data Mover. The problem occurred if specific bytes of a gidmap file were zero ( 2th, 3th and 512th bytes), resulting in the checksum being calculated incorrectly and the check during mount failing.

The data type for the calculation was corrected and the checksum is now calculated correctly.

377385 / 35859262

FileMover

The following entries describe fixed issues related to the Celerra FileMover:

260

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 While using FileMover, high memory utilization used in EDRS handles caused CIFS threads to hang and rendered the data unavailable.

The high memory utilization was because the memory used in EDRS read handles was not being freed quickly enough as they were getting blocked by another EDRS process. This issue was resolved by putting the close/free operations into their own thread pool.

385097 / 36947738

1 When having non-sequential tree IDs on the file system, renaming one of the directories representing one of the quota tree path (having an ID after the gap in the list) from the client did not reflect the change in the quota tree list using the nas_quotas command. The issue occurred because code was checking the sequential IDs only to the first gap.

Code has been fixed to check all existing tree IDs resulting in proper path display after a directory rename from the client.

386777 / 37133416

1 DHSM connections (using Kerberos for authentication) tried to re-use Kerberos tickets in the credential cache. The code used to check to see whether a ticket was available in the cache did not check to see if the ticket was expired, so the specified password was deleted. Subsequent code tried to use the expired ticket, then was unable to get a new ticket because the password had already been deleted, resulting in DHSM reconnection failure.

Added code (in Kerberos) to check for expired ticket in credential cache, and to delete the ticket if it is expired.

375592 / 35601816

2 For XML API clients, a query was made to get Shares with a filter of name="value", but the value field was specified as case-insensitive. Code enforced that values had to be case -sensitive, which caused older queries by XML API clients to fail if they did not specify the value correctly.

The C3 code was modified to accept case in-sensitive values for share names in order to remain backward compatible.

384012

261

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 User could not setup a NetApp server as a secondary server for DHSM. Netapp always returned unspecific error .

This has been fixed, and the CIFS client can now connect to NetApp server.

401210 / 37512728

Installation/Configuration/ Upgrades


Severity

Symptom Description

Fix Summary

Tracking Number

2 The sys_log showed the following repeatedly: Enclosure 0 failed to respond to "ping" at management switch A. These alerts were due to slow response from system Management Switches to Control Station ping requests.

Code was fixed to prevent false reboot messages and to harden some boxmonitor communication.

385933 / 36553396

1 If the partition table was corrupted, the PUHC did not capture the error and the upgrade failed at step "Create partitions for upgrade".

A new PUHC check was added to ensure that the disks are consistent so that the issue could be captured prior to the start of the upgrade.

391103 / 37606144

1 Control Station upgrade failed due to insufficient space on the "root" file system. In PUHC, there was no check available for the secondary Control Station (CS1) to verify the "available space" on "root" file system before upgrading.

Added new check for secondary CS to verify whether root file system have enough space prior to upgrade.

375108

262

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 Post Upgrade health check failed with error: Error HC_SYS_14505213994: "nas_storage -check" command timed out.

Code has been modified to avoid this timeout condition. The checks "cs::check_if_Symapi_data_is_pres ent" and "cs::check_if_Symapi_is_synced_with_Storage_System" are shifted to the end of the "cs" checks in checks.txt file to gain some time.

381208

1 An install of version 6.0.36.4 in an NSG2 system failed with the following: cmd: /nasmcd/sbin/cs_dhcpd_monitor -single_execution output: Unable to start cs_dhcpd_monitor -single_execution, already running!

This error is no longer displayed.

381753

2 If user's PATH variable was missing and they tried to run clariion_mgmt, the iptables file was not updated. This caused the following error during PAHC/PUHC: Storage System : Check SPA SPB communication Error HC_BE_14505017473: At least one iptables rule is missing. Action : Run "spa_spb_comm -setup" to fix the problem.

Adding full path to commands in clariion_mgmt using iptables-save to prevent this condition.

384764

1 While performing an out-of-family upgrade from 6.0.x to 7.0.x, a VG8 system was incorrectly reported as VG2.

The system now reports the correct model.

402851

2 During an in-family upgrade of a dual Control Station system, on the secondary CS, the task 'create IDE cache' was not run , therefore the files were copied directly to the common cache and wbem directories, thus filling up the Control Station root.

The task 'create IDE cache' is now run during upgrades to prevent this condition.

380619

1 In an out-of-family upgrade from 5.6 to 6.0, if the NAS service failed to stop

A fix has been provided to ensure that

389477

263

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

properly and some NAS file systems were left mounted at the end of the task. Upon Control Station reboot, the upgrade could not complete since the nas file systems were mounted to restore pre-reboot system state. Error - Partition table on /dev/hda is inconsistent.

no NAS related file systems are mounted by the upgrade for state restoration once the NAS service has been stopped during the upgrade.

2 After installing new language packs on top of an existing language pack in Unisphere, error messages were returned in English.

With this fix, error messages are returned in the installed language.

395453

2 When the file /nas/sys/nas_install_version did not exist, the upgrade failed with the following error: /nas/sys/nas_install_version: No such file or directory. There was no reference to which version of NAS code the system was originally installed with.

Added /nas/sys/nas_install_version into nasdb_backup code.

387041

iSCSI

The following entries describe fixed iSCSI-related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

264

Version 6.0.41.3


1 In a cluster environment, the iSCSI server hung and stopped responding. Host A had heavy I/O traffic to a LUN, and Host B tried to reserve the same LUN, using the SCSI-3 Persistent Reservation command. The reserve command hung, causing a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../dskdump.cxx at line: 1836 : ESX iSCSI access issue. The stack trace showed:

0x1f055ae0: 0x140ce4 waitForReboot+0x90 0x1f055b00: 0x140f69 fault_dump+0x67 0x1f055b20: 0x140e51 PANIC+0x29 0x1f055b30: 0xa2da32 _Z11paniccfgcmdR6Option+0x17e 0x1f055c70: 0x1c9f67 _ZN11cfgcmd_desc6invokeEPKcPP8CCMD_Msgii +0x1a7 0x1f055ce0: 0x2ae4a2 _Z10xml_cfgcmdP7macDataRP8CCMD_Msg+0x392 0x1f055d30: 0x2a9418 _ZN17macCCMDCommand_DB6invokeEPKcP7macDa taRP8C 0x1f055d60: 0x2a991e _Z17mac_CCMDDocParserP7macDataRP8CCMD_Ms g+0x20

The problem was caused by coding bug which created a small unprotected windows where an I/O completion callback was skipped. This was corrected and the unprotected window has been eliminated.

377382

NDMP

The following entries describe fixed issues related to Network Data Management Protocol and/or Backup:

Severity

Symptom Description

Fix Summary

Tracking Number

2 McAfee antivirus applications reset the last accessed time (atime) of a file after a clean scan. The Data Mover code updated the ctime when the atime was reset which impacted other features such as incremental NDMP backups.

The ctime is no longer updated as a result of an antivirus client resetting the atime.

387236 / 36692786

1 A race condition could occur during NDMP backup, when abort is called ,

Code has been fixed by using mutex lock to

395387 / 38052144

265

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

which results in a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 1d32a4 Err code: 0 Target addr: 0 *. DART stack trace:

0x2c4ba44: 0x14124c waitForReboot+0x90 0x2c4ba64: 0x1414d5 fault_dump+0x67 0x2c4ba84: 0x13a398 handle_fatal_fault+0xa 0x2c4baf0: 0x11fa5cd _ZN9NdmpdData7abortOpEv+0x253 0xf80e7ae0: 0x11fb42c _ZN16NdmpdDataAbortMH14messageHandlerEPv +0x94 0xf80e7b10: 0x1223aec _ZN14NdmpConnection15processMessagesEi+0 x2c6 0xf80e7b40: 0x122275c _ZN14NdmpConnection15processRequestsEv+0 x14 0xf80e7b60: 0x11f29c3 _ZN21ConnectionFileHandler11handleEventE v+0x1

protect the Data Mover from panicking under this condition

266

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 At the end of NDMP backup/restore, two threads accessed the same object during cleanup. One thread was reset while the other was using it. This led to a Data Mover panic with the following Dart/panic fault message: *** Page Fault Interrupt. Virt ADDRESS: 9eed24 Err code: 2 Target addr: ce **. The stack trace showed:

0x2c3ab24: 0x14124c waitForReboot+0x90 0x2c3ab44: 0x1414d5 fault_dump+0x67 0x2c3ab64: 0x13a398 handle_fatal_fault+0xa 0x2c3ac50: 0x9ee354 _ZN13PaxDataModuleD0Ev+0x4c 0x37a91c30: 0x11f3192 _ZN22NdmpdDataModuleManager10deallocateEP1 5Nd 0x37a91c40: 0x11f145e _ZN9NdmpdData16deallocateModuleEP15NdmpdDa taM 0x37a91c50: 0x11ede7e _ZN9NdmpdDataD1Ev+0x34 0x37a91c70: 0x12020e9 _ZN12NdmpdSessionD1Ev+0x63

Code now protects the cleanup during NDMP backup/restore to prevent this condition.

380146 / 36341620

2 Running security scanner against NDMP connections using port 10000 and having connection closed not cleanly, left that connection open on Data Mover. Eventually, when 2048 such connection will be left open, that will result in the following server error: NDMP: 3: Active NDMP session number reaches the limit 2048, No more NDMP sessions can be started. Contact your Customer Support representative and any NDMP backup will fail from this point.

Code has been fix to properly close the NDMP connection when network connection closed abnormally eliminating the problem.

380377 / 36311850

1 A directory read by an NDMP backup triggered blocks to be flushed to the file system, which in turn caused the changes to be snapped to the SavVol. The oldest

Code now allows the NDMP readdir operation to be aborted

375343 / 35697814

267

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

checkpoint had to be inactivated because the SavVol was full. However, the lock held by the NDMP backup on the file system prevented the unmount of the oldest checkpoint. This deadlock persisted until the panic was forced by the pending I/O. DART panic/fault message: >>PANIC: I/O not progressing LastVol touched Sn188 Kind 5 (ptr=0xace4c404). The stack trace showed:

0x6db1e90: 0x13fab4 waitForReboot+0x90 0x6db1eb0: 0x13fcb1 fault_dump+0x67 0x6db1ed0: 0x13fbb1 PANIC+0x29 0x6db1ee0: 0x822a30 _ZN9RawIO_IRP13checkProgressEv+0x11a 0x6db1fd0: 0x82290c _ZN17IrpProgressThread5startEv+0x6 0x6db1fe0: 0x14424c Sthread_startThread_internal+0xe

and retried when the checkpoint operation has completed.

268

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 During NDMP backup, in some rare conditions, a tape error led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../include/stream.hxx at line: 979 : datab::freeRef: db_ref <= 0 . The stack trace showed:

0xbccc5e50: 0x13f288 waitForReboot+0x90 0xbccc5e70: 0x13f485 fault_dump+0x67 0xbccc5e90: 0x13f385 PANIC+0x29 0xbccc5ea0: 0x1b407a _ZN5datab7freeRefEv+0x32 0xbccc5eb0: 0x96dc41 _ZN11MsgLinkList5EmptyEv+0x71 0xbccc5ed0: 0x970a76 _ZN9rw_thread5startEv+0xcb0 0xbccc5fe0: 0x143a08 Sthread_startThread_internal+0xe

Code was modified to properly handle media errors and prevent this panic.

376319

1 If the user created an NDMP account with text format, and put letter "A" into the account comment part, when DMA sent NDMP_CONNECT message with MD5 format, the decryption algorithm caused a Data Mover panic because the dividend became 0. DART panic/fault message: *** Divide exception. Virt ADDRESS: 11c036b ***. The stack trace showed:

0x2baf8ec: 0x140cec waitForReboot+0x90 0x2baf90c: 0x140f71 fault_dump+0x67 0x2baf92c: 0x139eec handle_fatal_fault+0xa 0x2baf9a0: 0x11c006e _ZN18NdmpdConnectAuthMH14messageHandlerE Pv+0x4 0xef84bb10: 0x11ec6f2 _ZN14NdmpConnection15processM

This is a wrong configuration. Code now handles this exception correctly and returns INVALID_ARGUMENT error.

390481 / 37563732

269

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

essagesEi+0 x2ba 0xef84bb40: 0x11eb388 _ZN14NdmpConnection15processRequestsEv+0 x14 0xef84bb60: 0x11bbee3 _ZN21ConnectionFileHandler11handleEventE v+0x1 0xef84bb80: 0x11ed956 _ZN19NdmpEventDispatcher8dispatchEim+0x7 ce

Network

The following entries describe fixed network-related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

1 A PCI bus failure caused the Neterion 10G NIC device to stop working. Undetected hardware failure of the Neterion 10G device resulted in an unusable port.

Code was modified to periodically check the hardware state of the Neterion device, and issue restart command if the device stops responding to I/O requests.

385326 / 36986690

270

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 When using virtual ethernet comprised of multiple devices (in this case configured an FSN over two 4-port trunks), memory allocation triggered a watchdog panic during power up. DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 840 : SYSTEM WATCHDOG. The stack trace showed:

0x20141ae4: 0x13fb94 waitForReboot+0x90 0x20141b04: 0x13fda9 fault_dump+0x67 0x20141b24: 0x13fc91 PANIC+0x29 0x20141b34: 0x141b41 system_watchdog+0x77 0x20141b44: 0x138966 _t_nmi+0x68 0x20141ba0: 0x30a072 _Z7LM_HaltP16_LM_DEVICE_BLOCK+0x1ea 0x20141bd0: 0x307b20 _Z15LM_ResetAdapterP16_LM_DEVICE_BLOCK+0 x48 0x20141c00: 0x306392 _Z20LM_InitializeAdapterP16_LM_DEVICE_BL OCK+0x

Code now separates the start into individual threads, causing the RT lock to be released between each device's start processing, thus avoiding the watchdog panic.

382462 / 36615128

2 User reported slow response time when mapping a Data Mover CIFS share or when mounting an NFS export. This issue only occurred when the Data Mover DNS client was explicitly configured with the TCP protocol (server_dns configures UDP by default). The DNS client used for discovering Domain Controllers and query for hostname addresses did not properly interpret two DNS responses included in the same TCP packet.

The DNS client code was modified to properly handle two DNS responses received in a same TCP packet.

399644 / 38312216

1 In some instances, the device "opaque" fields (ones that the device is supposed to copy unmodified) experienced single bit corruptions. DART panic/fault message: *** Page Fault Interrupt. Virt

The fix is to replicate the information so that it can be quickly verified when received.

377651 / 36005286

271

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

ADDRESS: 322e53 Err code: 0 Target addr: c **. The stack trace showed:

0x2ba3de4: 0x140ce4 waitForReboot+0x90 0x2ba3e04: 0x140f69 fault_dump+0x67 0x2ba3e24: 0x139ee4 handle_fatal_fault+0xa 0x2ba3e90: 0x31c7a8 _Z20LM_ServiceInterruptsP16_LM_DEVICE_BL OCK+0x1 0x4e81f80: 0x30fecf _ZN12bcmnicDevice5bcmRTEiPv+0x45 0x4e81fb0: 0x14226f _ZN15SchedRT_Initial5startEv+0xf3 0x4e81fe0: 0x1455cc Sthread_startThread_internal+0xe

1 Under heavy load, the Neterion 10G NIC hung, which led to loss of CIFS/NFS access, and logged many errors similar to the following in the server log: 2010-07-07 21:43:00: DRIVERS: 6: fxg-1-0 : Rescheduling RT.

Enhanced recovery mechanism to fix the issue on Neterion 10G with faulted hardware.

375769 / 35590150

272

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 In a rare case, during the disconnection of an SSL session, a certain sequence of network packets in the SSL module stream caused a Data Mover panic with the following: DART panic/fault message: Assertion failure: '_RDp != __null' on line 695 in file '../../dart_ssl/ssl_str.cxx'>>PANIC in file: ../assert.c at line: 26 : assert failed. The stack trace showed:

0x77ddea0: 0x14a464 waitForReboot+0x90 0x77ddec0: 0x14a703 fault_dump+0x67 0x77ddee0: 0x14a5d5 PANIC+0x29 0x77ddef0: 0x23de29 _assert+0x51 0x77ddf10: 0x11c9a4b _ZN10ssl_strCtx7wputSslEv+0x37 0x77ddf60: 0x11c9e99 _Z11sslstr_wsrvP5queue+0x4d 0x77ddf80: 0x20b593 _Z10STRserviceiPv+0x1af 0x77ddfa0: 0x14bba6 _ZN15SchedRT_Initial5startEv+0x132

SSL module stream function now handles the sequence of network packets correctly.

386273

NFS


Severity

Symptom Description

Fix Summary

Tracking Number

2 Write performance was reduced when file system free space was limited.

Enhanced the search for free space by selecting cylinder groups by their availability instead of a predefined order.

388304

273

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 When a file system was mounted with option “ntcredential”, the UNIX command chgrp failed on objects with owners that were members of more than 16 groups. The Data Mover took into account only groups that were part of NFS credentials, ignoring other groups specified in NT credentials.

The Data Mover now takes into account both NFS credentials and NT credentials when file system is mounted with option “ntcredential” and parameter cifs.acl.extendExtraGid is set. The limit of total number of groups is defined by parameter security.maxgroups.

385026

1 File system performance dropped when the files in the file system were not allocated correctly.

Modified the way files are allocated to optimize performance.

397730

1 When an NFSv3 client notifies the Data Mover that it has just rebooted, the Data Mover must revoke all the locks owned by the previous instance of this client. The removal of these locks triggered a system watchdog panic when the count of locks was huge (30K). DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 846 : SYSTEM WATCHDOG. The stack trace showed:

0x5085f24: 0x141254 waitForReboot+0x90 0x5085f44: 0x1414dd fault_dump+0x67 0x5085f64: 0x1413c5 PANIC+0x29 0x5085f74: 0x1432d1 system_watchdog+0x77 0x5085f84: 0x139f72 _t_nmi+0x68 0x5085fe0: 0x145bd6 Sthread_startThread_internal+0xe

Made the removal of these locks faster, so that the system watchdog is no longer triggered under this condition.

401962 / 38796804

274

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 If an NFS export was created with read-only host ip as 99.99.99.99, creating the NFS export failed with the error message: Modify NFS export server_2:/sanity_f123. server_2 : Invalid argument Export Error: Unknown option details. When the same process was done for Access hosts, it created the NFS export successfully.

Code was modified to ensure that exporting a path as read-only to specified hosts work correctly.

393021

2 User could not set permissions via NFSv4 clients on files and folders with ACL containing ACE of Windows local SYSTEM account. The following error appeared in server_log: NFS: 3: nfs4_checkSpecialAce: SID S-1-5-12 cannot be translated to NFSV4 : ACL can't be modified.

A new parameter nfsv4.ntAllSpecialUsers was added. When this parameter is set to 0 (default), Windows special users that are not listed in the RFC 3530 are returned as 'nobody' and changing permissions on objects with ACL containing local SYSTEM account will fail. When this parameter is set to 1, Windows special users are translated to a unique identifier, this allows modification of ACL for such objects from NFSv4 clients.

387582

1 During NFSv4 export database initialization, the server panicked with the following DART panic message: *** Page Fault Interrupt. Virt ADDRESS: 0x000179fb2e Err code: 0 Target addr: 0x0000000078 **

Code has been modified and a better mutex protection has been added to export database initialization.

396771

1 An NFSv4 client was sending a request to confirm the client ID to the server. In some cases, the Data Mover had already confirmed the ID, and didn’t know how to handle this situation. The Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 631ae3

Code has been fixed to handle this request by returning the following error message to the client: ID already in use and the Data Mover no longer panics in this situation.

384098 / 36843082

275

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

Err code: 0 Target addr: 28 **. The stack trace showed: 0x2acfb34: 0x13fb94 waitForReboot+0x90 0x2acfb54: 0x13fda9 fault_dump+0x67 0x2acfb74: 0x138d94 handle_fatal_fault+0xa 0x2acfbc0: 0x62f76c _ZN16NFSv4_ClientList7ConfirmEyPA8_KhR8n fs4_ctx 0x433fdcb0: 0x62822b _Z24nfs4_setclientid_confirmRP4msgbS1_RP hR8nfs 0x433fdd10: 0x622de0 _Z23nfs4_convertNfsv3Statusm+0x656 0x433fde90: 0x657949 _ZN13NfsGssRequest8dispatchEP4msgbPvP17A ccess_ 0x433fdee0: 0x605e16 _ZN17nfs_serverInitial5startEv+0x5d0

276

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 During nfsv3_lookup, if the path name length field in nfs header was corrupted, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 5e5e39 Err code: 0 Target addr: ffee8693 **.The stack trace showed:

0x28a6c64: 0x13fa98 waitForReboot+0x90 0x28a6c84: 0x13fc95 fault_dump+0x67 0x28a6ca4: 0x138cfc handle_fatal_fault+0xa 0x28a6e20: 0x5e2436 _ZN17nfs_serverInitial5startEv+0xac2 0xdefd1fe0: 0x144230 Sthread_startThread_internal+0xe

To prevent this panic, string length verification was added to verify the length field of NFS header of the request received from the network.

385131 / 36960608

Replication

The following entries describe fixed problems encountered while using Celerra Replicator, SnapSure or Checkpoints:

Severity

Symptom Description

Fix Summary

Tracking Number

1 A RepV2 session could not be stopped when the transfer was waiting for the destination SavVol to be extended.

Added a check for the stop request to continue during SavVol extension. This will allow a stop to proceed when the session is waiting for the destination SavVol to be extended.

387170 / 37192564

1 Repv2 suspends the writes to the destination file system when the SavVol is being extended. However, if the current SavVol usage value was equal to the

Code was fixed to prevent the checkpoint on the destination file system from becoming

388664 / 37379978

277

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

HWM (high watermark), writes were not suspended as expected. This caused the destination checkpoints to become inactive and if the file system on the destination was being accessed, an inconsistent view of the file system led to a Data Mover panic with the following DART panic/fault message: mangled directory ent. The stack trace showed:

0xcfdd56c0: 0x14124c waitForReboot+0x90 0xcfdd56e0: 0x1414d5 fault_dump+0x67 0xcfdd5700: 0x1413bd PANIC+0x29 0xcfdd5710: 0x778757 _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0xcfdd58d0: 0x7e272c _ZN9UFS_Vnode6dirbadEmPKc+0xea 0xcfdd5990: 0x7c0ee7 _ZN20UFS_DirHashInfo_dir3C1EP11Local_Vno de+0x5 0xcfdd5a10: 0x7e5b04 _ZN15UFS_DirOps_dir311initDirHashEP9UFS_ Vnode+ 0xcfdd5a40: 0x7d6ef6 _ZN15UFS_DirOps_dir37scandirEP9UFS_Vnode PK18Fi

inactive under this condition.

1 When using NFSv4, users were unable to copy directories on Data Mover shares from AIX clients using relative pathnames (for example, using “cp –r” command). The following error was reported by the client: No such file or directory. The issue occurred due to incorrect MOUNT_FILEID value returned to AIX client in NFSv4 request.

Data Mover now returns the correct MOUNT_FILEID value to NFSv4 clients allowing successful directory copies.

389170 / 37129124

278

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 Replication recovery tasks were not marked as complete if the destination Data Mover could not communicate with the source Data Mover during a VDM replication failover. A subsequent reboot of the Data Mover failed to mount the root file system of the VDM due to the presence of the incomplete tasks.

The code now checks whether a remount of the file system is needed during taskManager recovery, remounts the required file systems only, and correctly updates the status of the tasks.

394547 / 37882060

1 When using writable checkpoints and when un-cached I/O was used (in cases such as with VMWare/VAAI), Data Mover panicked with the following DART panic/fault message: >>PANIC: I/O not progressing LastVol touched Sh148108 Kind 8 (ptr=0x57d42204). The stack trace showed:

0x7c3de90: 0x14124c waitForReboot+0x90 0x7c3deb0: 0x1414d5 fault_dump+0x67 0x7c3ded0: 0x1413bd PANIC+0x29 0x7c3dee0: 0x8ad1b6 _ZN9RawIO_IRP13checkProgressEv+0x11a 0x7c3dfd0: 0x8ad092 _ZN17IrpProgressThread5startEv+0x6 0x7c3dfe0: 0x145bfa Sthread_startThread_internal+0xe

Code has been modified to handle the un-aligned 8KB Read/Write I/Os to writable checkpoints correctly.

386603

1 When the virtually provisioning options were modified on a file system replicated with Repv2, the replication database was not updated with the new settings. Therefore, after the Data Mover was rebooted, the old settings were applied to the file system. This resulted in the virtually provisioned file system not being extended.

The nas_fs - modify command now modifies the mount options correctly.

384264 / 36813726

279

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

2 The nas_replicate -stop and -delete operations required the user to enter a "mode" (set to both, source, or destination).

The nas_replicate command was enhanced so that the -stop and -delete options default to 'both'.

115393

2 When creating a checkpoint schedule, if a user changed the relative checkpoint from time based to named and then tried to mount the checkpoint, the checkpoint was mounted with the time based schedule instead on the named schedule.

Code has been fixed to use the correct schedule for mounting the checkpoint.

378004

280

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 During the merge of an empty checkpoint into an older checkpoint, after all the needed mappings were moved and the stableEntry was 2048*N. After a reboot, the wrong table of contents (TOC) buffer was flushed and the TOC became corrupt. This led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../DeltaVol.cxx at line: 7550 : Value in TOC is different from the value in btree. The stack trace showed:

0x9cb85df0: 0x14124c waitForReboot+0x90 0x9cb85e10: 0x1414d5 fault_dump+0x67 0x9cb85e30: 0x1413bd PANIC+0x29 0x9cb85e40: 0xfb1e7c _ZN19PoolManagedDeltaVol14verifyTOCValue Emy+0x 0x9cb85ea0: 0xfb1805 _ZN19PoolManagedDeltaVol10verifyTOCsEym1 1Verif 0x9cb85f00: 0xfb14fe _ZN19PoolManagedDeltaVol5mergeERP5Event+ 0xcf2 0x9cb85f80: 0xfc6d3c _ZN11mergeThread5startEv+0x1b4 0x9cb85fe0: 0x145bfa Sthread_startThread_internal+0xe

Code has been fixed to avoid the panic and to make the TOC pointer point to the right TOC buffer to be flushed.

391982 / 37718380

3 The nas_replicate-info command always shows the "Next Transfer Size" field to have the same value as "Current Transfer Size" field.

The Next Transfer Size (KB) field in the output of command nas_replicate -l was removed.

370770 / 34923460

1 During 1-to-Many replication sessions, empty signatures check prevented the source VDM from being changed to mounted state.

Code now skips the empty signatures check during one-to- many replication session to avoid this

386277

281

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

issue.

1 A deadlock condition occurred when RDE layer queries for SavVol stats while a checkpoint restore is in progress. The deadlock occurs during the dereference process in the DpLayer.

Code no longer requires the RDE thread to need a version set lock during the dereference process, hence avoiding a deadlock.

395215

1 If an unknown client (e.g. a port scanner) connected to RCP 8888 port too fast, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0 Err code: 0 Target addr: 0 ** . The stack trace showed:

0x2ba3d10: 0x140ce4 waitForReboot+0x90 0x2ba3d30: 0x140f69 fault_dump+0x67 0x2ba3d50: 0x139ee4 handle_fatal_fault+0xa 0x2ba3db0: 0x1d10d4 CloseStream+0x1b6 0x1ff3deb0: 0x12ce0b8 _ZN18RCP_SessionManager6acceptEPvP11sock addr_ 0x1ff3df40: 0x12ca690 _ZN10RCP_Server5startEv+0x3e6 0x1ff3dfe0: 0x1455cc Sthread_startThread_internal+0xe

When unknown clients attempt to connect to port 8888, the sessions are now properly closed and cleaned up.

384423 / 36899914

1 Replication transfer did not make progress as the SavVol on the destination could not be extended. User did not have enough information to understand why replication transfer did not progress.

An event is now posted to inform the user of the SavVol extension progress.

380166 / 36340862

282

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A checkpoint was stuck at Frozen state when the following three conditions occurred simultaneously: a PFS was being extended, checkpoint was being refreshed and SavVol was full. This led to blocking of snap IOs and triggered the following: DART panic/fault message: >>PANIC: I/O not progressing LastVol touched not Known

The stack trace showed: 0xdfe5df0: 0xc6c7bd _ZN9RawIO_IRP13checkProgressEv+0x3e9 0xdfe5fc0: 0xc6cb3e _ZN17IrpProgressThread5startEv+0x4 0xdfe5fd0: 0x16ad51 Sthread_startThread_internal+0x

Code was modified to ensure that when there are pending unmounting events for a checkpoint, the initFS allows mounting. This prevents the failure of checkpoint thaw for potential on-going checkpoint refresh operation.

373743

2 There was potential for slower than expected network transmit of RepV2 data when manually setting RCP TCP params.

Corrected an arithmetic error, and the decision logic for when to change the network transmit window.

381470 / 35658014

1 When a failover/switchover is executed on a replicationV2 session followed by a start in the reverse direction, the session will be set for manual refresh regardless of original "Max Out of Sync Time" setting.

The "Max Out of Sync Time" setting is now preserved and will be applied to the session when it is restarted after a failover/switchover.

389459

1 While replicating one Celerra to two other Celerras, the Data Mover on one of the destination Celerras did not clean up the connections to port 5085 and eventually the server_log became full of Too many connections on ports 5085 queue. The Data Mover on the destination file system panicked with the following DART panic/fault message: >>PANIC in file: ../dskdump.cxx at line: 1836 : Too many connections on

In the HTTPS server of the Data Mover, there was no timeout during the establishment of a SSL connection (between the acceptance of the TCP connection and the first packet sent by the client). The Data Mover was waiting for

381594 / 36464168

283

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

port 5085. The stack trace showed:

0x1facbae0: 0x141254 waitForReboot+0x90 0x1facbb00: 0x1414dd fault_dump+0x67 0x1facbb20: 0x1413c5 PANIC+0x29 0x1facbb30: 0xa44c58 _Z11paniccfgcmdR6Option+0x178 0x1facbc70: 0x1cbb8b _ZN11cfgcmd_desc6invokeEPKcPP8CCMD_Msgii +0x1a7 0x1facbce0: 0x2b2b3e _Z10xml_cfgcmdP7macDataRP8CCMD_Msg+0x392 0x1facbd30: 0x2adaf0 _ZN17macCCMDCommand_DB6invokeEPKcP7macDa taRP8C 0x1facbd60: 0x2adff6 _Z17mac_CCMDDocParserP7macDataRP8CCMD_Ms g+0x20

this first packet (the SSL clientHello), and until this packet was received, the HTTPS didn't process any new connection request. The HTTPS server was blocked and unavailable. This has been fixed.

1 After rebooting, the Data Mover would hang during the checkpoint recovery process and remain in "Loaded" state. Data Mover failovers would also fail.

A deadlock occurred while trying to correct a rare form of checkpoint corruption. The code now unlocks locked memory correctly, avoiding the deadlock.

378406 / 35911666

284

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A function in the code was consuming stack memory, which caused a stack overflow and led to a Data Mover panic during replication. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 1fa53e Err code: 0 Target addr: 0 **. The stack trace showed:

0x2acb158: 0x13fb94 waitForReboot+0x90 0x2acb178: 0x13fda9 fault_dump+0x67 0x2acb198: 0x138d94 handle_fatal_fault+0xa 0x2acb1c0: 0x56e267 krb5_cc_resolve+0x17 0x1eca92b0: 0x56e609 krb5_cc_default_suffix+0xa3 0x1eca92e0: 0x53e1db _Z26krb5_gss_release_principalPjPPv+0x46b 0x1eca93a0: 0x53ed90 _Z25krb5_gss_acquire_cred_extPjPv22gss_b uffer_ 0x1eca9400: 0x5456ff gss_acquire_cred_ext+0x4b

Code was modified to reduce memory consumption on the stack by using heap memory instead, preventing this condition.

383291 / 36431708

1 When a checkpoint was being refreshed or deleted the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../DeltaVol.cxx at line: 7550 : Value in TOC is different from the value in btree. The stack trace showed:

0x2971dd70: 0x141254 waitForReboot+0x90 0x2971dd90: 0x1414dd fault_dump+0x67

0x2971ddb0: 0x1413c5 PANIC+0x29 0x2971ddc0: 0xfaae5e _ZN19PoolManagedDeltaVol14verifyTOCValue Emy+0x

The checkpoint merge process now correctly handles the case when the merge has been interrupted by a Data Mover reboot.

381357

285

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

0x2971de20: 0xfaa7e7 _ZN19PoolManagedDeltaVol10verifyTOCsEym1 1Verif 0x2971de80: 0xfac73f _ZN19PoolManagedDeltaVol11changeToTOCEv+ 0x8b 0x2971deb0: 0xfac110 _ZN19PoolManagedDeltaVol13findFirstFreeE RyRP5E 0x2971df00: 0xfa9c62 _ZN19PoolManagedDeltaVol5mergeERP5Event+0x 474

286

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 There was a race between SnapSure smart snap and delete command. Smartsnap traversed from the latest ckpt to oldest ckpt to decide whether a block is a smartsnap block, while at the same time, the oldest ckpt was deleted by user command. During smart snap, code needed to find mapping from the oldest ckpt and only found it was destructed, then panicked the Data Mover with the following DART panic/fault message: *** GP exception. Virt ADDRESS: b877dd. Err code: 0 *** . The stack trace showed:

0x172f458: 0x13db94 waitForReboot+0x90 0x172f478: 0x13dd91 fault_dump+0x67 0x172f498: 0x137d54 handle_fatal_fault+0xa 0x172f4e0: 0xb91b91 _ZN15snapshotContext10BlockInUseEP8Delta VolyP8S 0x242e9be0: 0xb7d6dc _ZN19PoolManagedDeltaVol18writeBlockFilt eredEP 0x242e9c90: 0xb8d232 _ZN10SnappedVol18writeBlockFilteredEP10D iskIO_ 0x242e9cb0: 0xb8d1ed _ZN10SnappedVol18writeBlockInternalEP10D iskIO_ 0x242e9cd0: 0x731731 _ZN11HyperVolume18writeBlockInternalEP10 DiskIO

This fix is to close this race by suspending smart snap if there is a delete operation in progress.

377960 / 35997088

1 The Data Mover experienced repeated Watchdog panics when an attempt was made to create the first ReplicationV2 session. This was because the Control Station to DART interconnect did not exist. DART Server183ic/fault message: >>Server183IC in file: ../sched.cxx at line: 846 : SYSTEM WATCHDOG. The stack

During a Data Mover reboot, a check will be performed to verify that the Control Station to DART interconnect is present. If the interconnect is missing it will be added by running "nas_cel -

371904 / 35235954

287

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

trace showed:

0x5085f24: 0x14124c waitForReboot+0x90 0x5085f44: 0x1414d5 fault_dump+0x67 0x5085f64: 0x1413bd Server183IC+0x29 0x5085f74: 0x1432c9 system_watchdog+0x77 0x5085f84: 0x13IP_ADDR_nmi+0x68 0x5085fe0: 0x145bfa Sthread_startThread_internal+0xe

update id=0".

288

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A SavVol was almost full, and the most recent ckpt was about to be inactivated to save space. A snap request on this ckpt happened just before the deactivation, which caused an invalid function in the code and led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../DeltaVol.cxx at line: 3792 : doSavVolIO failure. The stack trace showed:

0x236025e60: 0x1309df4 _ZN19PoolManagedDeltaVol10doSavVolIOEyjP ci16 0x236025eb0: 0x130ed59 _ZN19PoolManagedDeltaVol22FlushSnapShotL ogLo 0x236025ef0: 0x130f0c7 ZN19PoolManagedDeltaVol16FlushSnapShotLo gE1 0x236025f20: 0x12fcdf5 _ZN12IntentLog_CB8allocateEv+0x15 0x236025f40: 0x1348c65 _ZN21ckptBlockAllocatorThr5startEv+0x375 0x236025fd0: 0x15519f Sthread_startThread_internal+0x7

This issue was fixed by adding a hdr check condition to FlushSnapShotLogLocked, thus avoiding a panic.

370442

1 There was a race condition between ckpt restore complete routine and ckpt delete command. SavVol was full during ckpt restore process, it erased all the ckpts on the SavVol, while a ckpt on this SavVol was being deleted by user's command, it double-destructed the ckpt which led to a panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: f8d2d0 Err code: 0 Target addr: 0 **. The stack trace showed:

0x2ba3bd4: 0x140ce4 waitForReboot+0x90

A lock mechanism was added to avoid the race condition.

159333

289

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

0x2ba3bf4: 0x140f69 fault_dump+0x67 0x2ba3c14: 0x139ee4 handle_fatal_fault+0xa 0x2ba3c70: 0xf8cf26 _ZN19PoolManagedDeltaVol18ReleaseChunkBi tMapEv+ 0x1f453d60: 0xf7f332 _ZN19PoolManagedDeltaVol15ReleaseSnapsho tEv+0x 0x1f453dc0: 0xf7aa74 _ZN19PoolManagedDeltaVol19SetSnapShotInA ctiveE 0x1f453e40: 0xf94af2 _ZN15snapshotContext17BroadCastInactiveE PKcP8D 0x1f453e80: 0xf802a8 _ZN19PoolManagedDeltaVol11RestoreStopEv+ 0xa4

1 A replication stop/delete/failover operation task hung on the source side of replication, if the destination side was already stopped or deleted. The task indicated the following error: Error 13160415328: Registration request rejected.

Code has been modified to ensure replication commands such as stop/delete/failover no longer hang when the destination side is stopped or deleted.

380942

290

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 A Data Mover hung when an SMB thread tried to close and compress a file on a file system to which a replication thread was trying to unmount. DART panic/fault message: >>PANIC in file: ../dskdump.cxx at line: 1836 : cifs block. The stack trace showed:

0x2128bae0: 0x14124c waitForReboot+0x90 0x2128bb00: 0x1414d5 fault_dump+0x67 0x2128bb20: 0x1413bd PANIC+0x29 0x2128bb30: 0xa450c8 _Z11paniccfgcmdR6Option+0x178 0x2128bc70: 0x1cbdab _ZN11cfgcmd_desc6invokeEPKcPP8CCMD_Msgii +0x1a7 0x2128bce0: 0x2b205a _Z10xml_cfgcmdP7macDataRP8CCMD_Msg+0x392 0x2128bd30: 0x2ad00c _ZN17macCCMDCommand_DB6invokeEPKcP7macDa taRP8C 0x2128bd60: 0x2ad512 _Z17mac_CCMDDocParserP7macDataRP8CCMD_Ms g+0x20

Replication code now verifies that the file system is not locked by any other threads under such conditions before trying to unmount a file system.

388310 / 37319436

1 A bug which allowed a null value to be returned to SnapSure led to the following Data Mover panic: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: fa59c6 Err code: 0 Target addr: 4 **. The stack trace showed:

0x2c3b184: 0x14124c waitForReboot+0x90 0x2c3b1a4: 0x1414d5 fault_dump+0x67 0x2c3b1c4: 0x13a398 handle_fatal_fault+0xa 0x2c3b260: 0xfbabe0 _ZN10SnappedVol18writeBlockFi

The code has been fixed so that a NULL value is no longer returned, and no panic results.

382916 / 36634308

291

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

lteredEP10D iskIO_I 0xe9871240: 0xfbb091 _ZN10SnappedVol16aioInternalWriteEP10Dis kIO_IR 0xe98712a0: 0xfbaca8 _ZN10SnappedVol11aioInternalEP10DiskIO_I RPP8ai 0xe98712d0: 0x8c2361 _ZN11HyperVolume11aioInternalEP10DiskIO_ IRPP8a 0xe98713b0: 0x8d4007 _ZN17UnCachedIO_Object10lio_listioE7LioM odePP8

292

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 If an application label containing spaces was used to create an iscsi LUN replication, when user issued a reverse command, the Data Mover panicked with the following Assertion failure: 'dicSessionp != NULL' on line 931 in file '../DpTaskManager.cxx' >>PANIC in file: ../assert.c at line: 26 : assert failed. The stack trace showed:

0xe0031e30: 0x14124c waitForReboot+0x90 0xe0031e50: 0x1414d5 fault_dump+0x67 0xe0031e70: 0x1413bd PANIC+0x29 0xe0031e80: 0x1ff975 _assert+0x51 0xe0031ea0: 0x10c63c8 _ZN13DpTaskManager20createSyncDicSession ERP15 0xe0031ee0: 0x10cba57 _ZN33DpTaskManagerCopyReverseInitiator8s tartS 0xe0031f30: 0x10caa07 _ZN33DpTaskManagerCopyReverseInitiator14 start 0xe0031f50: 0x10c844e _ZN33DpTaskManagerCopyReverseInitiator14 Start

The application label is now validated and rejected if it contains spaces.

383407 / 36730552

Security

The following entries describe fixed Security related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

1 When simultaneous access by two threads to the same SSL certificate was attempted, the Data Mover panicked with the following DART panic/fault message:

Added a mutex lock to protect the access to the SSL certificate and

393878

293

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

>>PANIC in file: ../malloc.cxx at line: 351 : Memory corrupted. The stack trace showed: 0xe5fefc40: 0x140ce4 waitForReboot+0x90

0xe5fefc60: 0x140f69 fault_dump+0x67 0xe5fefc80: 0x140e51 PANIC+0x29 0xe5fefc90: 0x1fbdb7 free+0x69 0xe5fefcc0: 0xda5b38 CRYPTO_free+0xc 0xe5fefcd0: 0xd3fe2d BUF_MEM_free+0x2c 0xe5fefcf0: 0xd918ba X509_NAME_dup+0xe5 0xe5fefd10: 0xd8c7a1 ASN1_item_ex_free+0x171

prevent this condition.

2 The server_ldap command gave no indication whether the LDAP Service configuration succeeded. User needed to do additional queries or check the server logs to find out.

A new alert-level event was added for LDAP Service configuration failures. If a configuration fails, the following message will be in the sys_log and (if configured) emailed to the user: An attempt to configure the LDAP Service has failed. See the server log for applicable error messages. Correct the error, then retry the configuration command..

149215 / 29904638

294

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 In version 6.0, code was added to log an event if an LDAP service configuration attempt failed. However, the code was added to several error loops that included 20 second delays before retrying the configuration attempt. the result was that the 20 second delay is always run whether there is an error or not, causing a delay in the service configuration.

Fixed the error while loops to only run the 20 second delay if an error occurs.

385248

1 A race condition between periodic check for reconfiguration and manual reconfiguration may lead to an LDAP deadlock.

This issue occurred when performing continuous LDAP reconfigurations in a short period of time. The LDAP timer no longer signals the event if already signaled.

383944

SRDF

The following entries describe fixed Symmetrix Remote Data Facility related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

3 The fs_timefinder -M off command running against fs_group failed with the following error message. Error 13431668739: File system group has a file system which is backup of a Primary File System (PFS) which is actively mirroring to file system <fsname>.

Removed the check which caused this issue. The fs_timefinder -M off command with fs_group now works correctly.

371759

1 During a nas_rdf -restore, a gatekeeper lock condition should have forced the command to exit, the command was allowed to continue due to a unique condition. This caused the nas_rdf -restore command to hang.

The nas_rdf command will now exit when excessive lock contention is experienced. If a gatekeeper lock will not release, identify the lock and manually

382490 / 36173416

295

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

release it. Identify the pid holding the lock with the following command: /usr/symcli/bin/symcfg list -sem -v Release the lock with the following command: ipcrm -s <pid> Then rerun the nas_rdf -restore command.

Storage

The following entries describe fixed storage-related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

1 When the source storage system was modified after the destination had been activated, the nas_mview -restore command failed.

Enhancements to the nas_mview command have been applied to detect storage system changes allowing the -restore to complete successfully.

382625 / 36666666

296

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 When paths to a device failed, and some (administrative) action caused a disk signature to be verified, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x0000a43da9 Err code: 0 Target addr: 0x0000000010 **. The stack trace showed:

0x3680760: 0xbf6450 _ZN7DskDump9checkdumpEPKc+0x7a 0x36807a0: 0xbf665e dump_return+0x4 0x36807b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 0x36807d0: 0x14ce5d fault_dump+0x89 0x36807f0: 0x13378d handle_fault+0x1d 0x689fd20: 0xa4aec7 _ZN11BasicVolume21insertToIoQueueCommonE P8SgIO_ 0x689fd60: 0xa53518 _ZN11BasicVolume15IOBlockInternalEP10Dis kIO_IRP 0x689fdf0: 0xa57ab9 _ZN11BasicVolume18writeBlockInternalEP10 DiskIO_

Fixed a locking bug which caused stale queue data to be referenced.

376685

1 When using UDP-based applications with large xfer sizes or using TCP and network environmental conditions caused fragmentation to occur, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 0x00003cb317 Err code: 0 Target addr: 0x0000000006 **. The stack trace showed:

0x3680760: 0xbf6490 _ZN7DskDump9checkdumpEPKc+0x7a 0x36807a0: 0xbf669e dump_return+0x4

Updated code to correct issues with fragment list management and prevent this pan

388143 / 37288224

297

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

0x36807b0: 0x14c7d2 _ZN12Memory_Frame13setProtectionEm19Memo ry_Prot 0x36807d0: 0x14ce5d fault_dump+0x89 0x36807f0: 0x13378d handle_fault+0x1d 0x6880d90: 0x3c08e7 ip_mxrput+0x10d 0x6880dd0: 0x37b4bd _Z7bcmreadPvP4msgbtjj+0x4e7 0x6880ee0: 0x38b2b2 _Z20LM_ServiceInterruptsP16_LM_DEVICE_BL OCK+0x3

1 Periodically the Data Mover NIS client was unable to connect to the NIS server and switched to another NIS server. The server log reported an error like: The NIS client is unable to connect to the NIS server w.x.y.z : RPC failure.

In some instances the NIS server can be slow to respond so the NIS RPC retry timeout is increased to 5 seconds and the NIS RPC request timeout increased to 15 seconds. This will prevent switching to another NIS server when there is a brief delay in the response from the NIS server.

375293 / 34806928

1 User removed 5 LUNs from Celerra to change HLU’s (host LUN unit) and was unable to import them back even though no errors were logged. LUNs were part of a MirrorView/S consistency group and the command nas_mview -init needed to be used to diskmark them, but the Celerra error message gave no indication of this.

Checks were added in the code and error messages enhanced to determine what needs to be done to diskmark different configurations.

126624 / 25702698

298

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

1 The Data Mover received incorrect inquiry data from a backup tape device, which later triggered a panic due to dereferencing of a null volume pointer. The Data Mover panicked with the following DART panic/fault message: ** Page Fault Interrupt. Virt ADDRESS: 8b341f Err code: 0 Target addr: 0 **. The stack trace showed:

0x2c2b494: 0x141254 waitForReboot+0x90

0x2c2b4b4: 0x1414dd fault_dump+0x67 0x2c2b4d4: 0x13a3a0 handle_fatal_fault+0xa 0x2c2ba60: 0x296c17 _Z16macthr_sysconfigP4msgb+0x1b5f 0x21e25fa0: 0x27c52a _ZN11Mac_Initial5startEv+0x3e0

The fix avoids null-pointer dereferencing when invalid Inquiry data is returned from a device, preventing the panic.

372043 / 35302478

1 An attempt to specify more than 8 paths on a volume disk command resulted in a Data Mover panic with the following DART panic/fault message: *** GP exception. Virt ADDRESS: 13990eb. Err code: 0 ***. The stack trace showed:

0x2c4bcb8: 0x14124c waitForReboot+0x90 0x2c4bcd8: 0x1414d5 fault_dump+0x67 0x2c4bcf8: 0x13a398 handle_fatal_fault+0xa 0x2c4bd20: 0x1399061 _ZN14XmlElementNodeD0Ev+0x1d 0x25429cf0: 0x13991ca _ZN14XmlElementNode14removeChildrenEv+0x 2c 0x25429d10: 0x1399059 _ZN14XmlElementNodeD0Ev+0x15 0x25429d20: 0x13991ca

An attempt to specify more than 8 paths did not fail but caused memory corruption and a Data Mover panic. The code that parses volume disk commands has been modified to explicitly check and fail with the following error if more than 8 paths are specified: logIO::logmsg (LOG_STORAGE, LOG_ERR, 5825 logIO::logmsg (LOG_STORAGE, LOG_ERR "Cannot create Basic Volume %s, numpaths parameter invalid %d\n .

378239

299

Version 6.0.41.3


Severity

Symptom Description

Fix Summary

Tracking Number

_ZN14XmlElementNode14removeChildrenEv+0x 2c 0x25429d40: 0x1398fc1 _ZN14XmlElementNodeD2Ev+0x15

1 When performing a PFS restore from the oldest checkpoint, if the SavVol became full leading to oldest checkpoint inactivation, the Data Mover panicked with the following DART panic/fault message: >>PANIC: I/O not progressing LastVol touched D6853119_0 Kind 8 (ptr=0x115ed0038. The stack trace showed:

0xc3addf0: 0xc566bd _ZN9RawIO_IRP13checkProgressEv+0x3e9 0xc3adfc0: 0xc56a3e _ZN17IrpProgressThread5startEv+0x4 0xc3adfd0: 0x16b4b1 Sthread_startThread_internal+0x7

Code has been fixed to prevent the panic for this condition.

370887

Impact Level Severity 1 Symptom Desc. Celerra kept reporting the trespassing of HDDs from SPA to SPB.

Server log showed many CAM errors like this: 2010-01-11 13:11:42: 81874518019: CAM: 3: 3: Warning: The SCSI HBA 16 is inaccessible. 2010-01-11 13:11:42: 13155041280: CAM: 3: I/O Error: c16t0l0 Irp 0x57c25f04 CamStatus 0x05 ScsiStatus 0x08 Sense 0x05/0x04/0x00 2010-01-11 13:11:42: 13155041280: CAM: 3: I/O Error: c16t9l15 Irp 0x3c641104 CamStatus 0x05 ScsiStatus 0x08 Sense 0x05/0x04/0x00 and sys_log showed errors like this: Dec 8 23:40:28 2009:CS_PLATFORM:NaviEventMonitor:CRITICAL:101::::1260312028:Disk d8 has been trespassed. Dec 8 23:40:28

300

Version 6.0.41.3


2009:CS_PLATFORM:NaviEventMonitor:CRITICAL:101::::1260312028:Disk d11 has been trespassed. Dec 8 23:40:28 2009:CS_PLATFORM:NaviEventMonitor:CRITICAL:101::::1260312028:Disk d13 has been trespassed. Dec 8 23:40:28 2009:CS_PLATFORM:NaviEventMonitor:CRITICAL:101::::1260312028:Disk d14 has been trespassed.

Fix Enhanced the error status handling code to handle busy luns without forcing a trespass.

Service Request No. 31963984

UFS


Severity

Symptom Description

Fix Summary

Tracking Number

1 During the UTF-8 conversion process for a file system, if a Data Mover attempted to convert a directory owned by a user who was exceeding quotas, the Data Mover panicked with the following DART panic/fault message: Cannot read unnamed stream inode for container.

The stack trace showed: 0xdf591310: 0x14124c waitForReboot+0x90 0xdf591330: 0x1414d5 fault_dump+0x67 0xdf591350: 0x1413bd PANIC+0x29 0xdf591360: 0x778b27 _ZN14UFS_FileSystem28markCorruptedFsForU nmount 0xdf591520: 0x7cabba _ZN9UFS_Vnode18setUnnamedVnodePtrEv+0x192 0xdf591570: 0x777d74 _ZN14UFS_FileSystem16findExistingNodeEmR 11File 0xdf5915c0: 0x777976 _ZN14UFS_FileSystem8findNodeE

Code was updated to be more tolerant of this error, and will no longer panic under this condition. The following error is returned if the conversion fails: UFS_Vnode::setUnnamedVn odePtr -- fs <fsid> inum <inode number> Quota Exceeded\n where 'Quota Exceeded' can be user, group, tree quota, or just general file system out of space. The node being accessed will not be initialized, and the user will not be able to read the affected file/directory.

382495 / 36659580

301

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

mR11File_St atust8 0xdf591610: 0x7e12ea _ZN15UFS_DirOps_dir38getVnodeEP9UFS_Vnod emPP9V.

Version 6.0.40.8


The following are fixed problems related to the Celerra Management Applications, i.e., Unisphere, Celerra Startup Assistant, or Celerra MMC Snap-ins:

Severity

Symptom Description

Fix Summary

Tracking Number

1 User was unable to add or manage newer versions of Celerra systems to legacy domains in Unisphere due to restrictive version check.

Version check was modified to allow Celerra systems with newer version to be added to or managed in the domains of current version systems using Unisphere.

395097

2 Network statistics were unavailable through Celerra Manager when Celerra had 10 gigabit network devices. Graph always showed "Data Loading..." and graph options menu didn't include 10 gigabit network devices.

Added 10 gigabit network devices for inclusion in Celerra Manager network graphs.

379658 / 36124794

302

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

2 When a Celerra was in production mode, Unisphere took a considerable time to load. The loading symbol (rotating arrows) was persistently shown near the IP address of the Control Station in the Unisphere left pane.

An underlying command was used to find the model of the Celerra system which caused serious latency for the loading of the Celerra pages in Unisphere. The command model –x was modified to include caching of the model of the system which speeds up the Unisphere pages.

374949 / 35327164

2 Jserver fails to parse output of the server_cifs server_x -o audit command due to some extraneous lines coming from Data Mover response resulting in the process hanging on Control Station. Over time, lots of hanging server_cifs processes consumed all the memory on the Control Station and the Control Station had to be rebooted.

Code has been fixed to ensure that Jserver parsing code never fails on unexpected output lines and ensures that “server_cifs” processes spawned by Jserver always complete and prevent excessive memory utilization from occurring.

375192 / 35638732

CIFS


Severity

Symptom Description

Fix Summary

Tracking Number

1 The Windows group Policy was maintained on the domain controller in unicode files. If a user replaced the unicode file with a non-unicode security template, then the file was not read properly and if large enough (greater than 54Kbytes) the Data Mover panicked with the following DART panic/fault message: NIC in file: ../malloc.cxx at line: 525 :

Updated the GPO code to check if the file is unicode (first two bytes of the file are 0xFFFE), then process the file either as unicode or non-unicode, retrieving and maintaining all settings

384136 / 36849782

303

Version 6.0.40.8


Memory corrupted..

that are applicable to the Data Mover.

2 When implementing CAVA using a CIFS server with disjoint DNS and Windows domains, CAVA would not go online unless the default CIFS server was specified in viruschecker.conf.

Code now uses the FQDN of the first interface of the default CIFS server for the CIFS server.

365843 / 34585998

2 When doing a lot of CAVA /CEPA configurations while the CAVA service was stopped, the configuration was not freed and the streams used to send the heartbeat were not freed. This led to a memory leak which could result in a Data Mover panic.

Code now frees the configuration and the streams used to send the heartbeat when VC is stopped.

381971 / 36571200

304

Version 6.0.40.8


1 Under heavy load, when the CEPP feature was enabled, Data Mover panicked with the following due to a race condition between two CIFS threads: DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: c601c2 Err code: 0 Target addr: 0 **. The stack trace showed:

0x2c3bc54: 0x14124c waitForReboot+0x90 0x2c3bc74: 0x1414d5 fault_dump+0x67 0x2c3bc94: 0x13a398 handle_fatal_fault+0xa 0x2c3bcd0: 0xc5b803 _ZN8fid_info15attachCloseCeppEP9CIFS_CEP P+0x59 0xfa6ffca0: 0xc60ada _ZN9CIFS_CEPP17setDeferredActionEv+0x2a 0xfa6ffcc0: 0xc5262a _ZN8AppLibNT5closeEP8fid_infoiR11File_St atusP1 0xfa6ffe80: 0xbf8302 _ZN13smb_streamCtx16closeOpenedFilesEt+0 xd2 0xfa6ffed0: 0xb09df2 _ZN13smb_streamCtx10credRemoveER11cRefer encerI


385161 / 36978078

1 After upgrading to 6.0 release, when using NTLM authentication, users from trusted domains might no longer be able to access resources (server, share, file, directory) granted through Universal group membership.

Code now correctly processes extra SIDs returned by the Domain Controller.

385626 / 36963164

1 DHSM connections (using Kerberos for authentication) try to re-use Kerberos tickets in the credential cache. The code used to check to see whether a ticket was available in the cache did not check to see if the ticket was expired, so the specified password was deleted. Subsequent code tried to use the expired ticket, then was unable to get a new ticket because the password had already been deleted, resulting in DHSM reconnection failure.

Added code (in Kerberos) to check for expired ticket in credential cache, and to delete the ticket if it is expired.

375592 / 35601816

305

Version 6.0.40.8


DART panic/fault message: >>PANIC in file: ../dskdump.cxx at line: 1836 : DHSM recall issues. The stack trace showed:

0x21201ae0: 0x14124c waitForReboot+0x90

0x21201b00: 0x1414d5 fault_dump+0x67

0x21201b20: 0x1413bd PANIC+0x29

0x21201b30: 0xa44e08 _Z11paniccfgcmdR6Option+0x178

0x21201c70: 0x1cbdab _ZN11cfgcmd_desc6invokeEPKcPP8CCMD_Msgii +0x1a7

0x21201ce0: 0x2b205a _Z10xml_cfgcmdP7macDataRP8CCMD_Msg+0x392

0x21201d30: 0x2ad00c _ZN17macCCMDCommand_DB6invokeEPKcP7macData RP8C

0x21201d60: 0x2ad512 _Z17mac_CCMDDocParserP7macDataRP8CCMD_Ms g+0x20

1 When users in Active Directory were in

"lastname, firstname" format, the resulting distinguished name included an escaped comma; i.e., "\,". The Data Mover search filter processing for LDAP did not process escaped commas correctly, and returned incorrect search results.

Code was modified to allow an escaped comma to be included in the search filter string.

369461 / 3480317

306

Version 6.0.40.8


2 Data Mover stopped parsing nsswitch.conf file when encountering an invalid entry in it, the rest of the configuration lines were rejected and no error was reported about it which might prevent certain functionality from working, such as LDAP for CIFS UID/GID mapping.

Code has been fixed to parse the whole file and ignore invalid entries, but report an error message in the server_log about the invalid entry.

372383 / 35287884

1 User tried to disable file filtering temporarily through the cifs.enableFileFiltering parameter. The parameter description says it applies immediately but it was found that changing the value had no effect if the connection was still opened. An attempt by client to create file "\nasadmin\New Microsoft Word Document.doc" failed with STATUS_ACCESS_DENIED. The Data Mover allowed the write after the Data Mover was rebooted.

The parameter was modified to not require any user action (rebooting), even for connection active when the parameter has been changed.

373375 / 35092392

Control Station


Severity

Symptom Description

Fix Summary

Tracking Number

2 The Data Mover did not failover after the panic. This occurred because the code did not notify all the processes to release the process locks, and a checkpoint related process was holding a lock which prevented the failover.

Code now ensures that all processes are appropriately notified to exit and release the locks.

370795 / 35139496

2 Users observed the following error in the GUI alert log: Parsing error. Invalid XML. The message did not contain any specific information about the parsing error.

The function that logs the messages containing the reason for the parsing failure was altered to log the entire message with the specific reason of the parsing error in /nas/log/nas_log.al.mgmtd file. This

361847 / 33292418

307

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

message will contain the entire xml string, including in this case, the invalid characters.

2 When NTP services were started on a Data Mover with the interval greater than 1440 minutes specified, it was set on the Data Mover but then the Data Mover panicked with the following DART panic/fault message: *** Divide exception. Virt ADDRESS: a2edaf ***. The stack trace showed:

0x2ba3d0c: 0x140ce4 waitForReboot+0x90 0x2ba3d2c: 0x140f69 fault_dump+0x67 0x2ba3d4c: 0x139ee4 handle_fatal_fault+0xa 0x2ba3ed0: 0xa2fa0c _ZN8TimeSync5startEv+0xb2 0xfdd37fe0: 0x1455cc Sthread_startThread_internal+0xe

Code now prevents a value greater than 1440 minutes (or 24 hours) for NTP sync interval.

361893 / 34311052

CORE


Severity

Symptom Description

Fix Summary

Tracking Number

308

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 A rolling panic occurred when a local netgroup file contained a nested netgroup. Users had CIFS access, but no NFS access. DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 14dd7a Err code: 2 Target addr: 8f **. The stack trace showed:

0x2c2cce4: 0x141254 waitForReboot+0x90 0x2c2cd04: 0x1414dd fault_dump+0x67 0x2c2cd24: 0x13a3a0 handle_fatal_fault+0xa 0x2c2cd50: 0x1c54a3 _Z13CollectStreamPvS_+0x77 0x5081e30: 0x1d577a ReadPutq+0x2b2 0x5081e60: 0x6350eb nfs_rput+0xa47 0x5081eb0: 0x634b8c nfs_rput+0x4e8

Code now processes only ten levels of nesting in a netgroups. If there exist netgroups nested to a depth greater than ten, the following error message is now printed in the server log: NETGROUP netgElem::IsMember Looping on <netgroup name> (loop in nested netgroups) \n.

374681 / 35598996

1 When a Data Mover was mistakenly zoned with more than 32 other initiators, the Data Mover was unable to see the storage and failed to boot.

The Data Mover will now detect when too many devices are incorrectly zoned into the Data Mover and will log diagnostic information that will help quickly identify the issue. The following messages will appear in the server log under such conditions: Failed to add device, no resources and Failed to add device, unsupported login type

366770 / 34764834

309

Version 6.0.40.8


Deduplication


Severity

Symptom Description

Fix Summary

Tracking Number

1 On a system with heavy deduplication activity, the Data Mover did not release memory in the EDRS (EMC Data Reduction Services) portion of the code and caused a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../addrspac.cxx at line: 289 : Out of memory. The stack trace showed:

0xde21b0a0: 0x14124c waitForReboot+0x90 0xde21b0c0: 0x1414d5 fault_dump+0x67 0xde21b0e0: 0x1413bd PANIC+0x29 0xde21b0f0: 0x13bef6 _ZN19Memory_AddressSpace10allocPagesEi12 Memory 0xde21b110: 0x1fe726 _Z11more_memoryjP9free_list+0x4a 0xde21b130: 0x1fe7ec malloc+0xa0 0xde21b160: 0x152dbc2 csx_rt_os_mem_raw_alloc_maybe+0x14 0xde21b180: 0x152bc16 csx_rt_mem_raw_alloc_maybe+0x14

Code was fixed to correctly release memory and prevent this panic.

375789 / 35737678

1 Under rare condition, removing a deduplicated file, creating a checkpoint and performing a restore on the same file system at the same time, resulted in a deadlock due to lock interdependencies resulting in blocked threads on the Data Mover and eventually causing access issues.

Code has been fixed to properly handle interdependencies lock so no deadlock would occur under this condition.

372616 / 35375114

310

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 File systems with deduplication enabled (including NDMP backup checkpoint file systems) failed to unmount. Messages similar to the following were logged to the server log: 2010-07-28 19:24:11: DEDUPE: 3: openFileData failed: fsid 483 status RDE_STATUS_DECOMPRESSION_ERROR.

Code is updated to ensure files are closed when a DECOMPRESSION ERROR is detected.

374310 / 35478060

1 When a read of a dedupe file system resulted in sending a larger request than the fibre driver could process, the Data Mover panicked with following DART panic/fault message: >>PANIC in file: ../BVolumeIrp.cxx at line: 169 : Problem building SG list. Panicking Processor: 2. The stack trace showed:

ee2a90: 0xa9e735 _ZN8SgIO_IRP10initializeEP10DiskIO_IRPP11B asi ee2ad0: 0xa54b3c _ZN11BasicVolume11aioInternalEP10DiskIO_IR PP8 ee2ba0: 0xa55cd7 _ZN11HyperVolume11aioInternalEP10DiskIO_IR PP8 ee2c70: 0xa5f213 _ZN11SliceVolume11aioInternalEP10DiskIO_IR PP8 ee2ca0: 0xa55cd7 _ZN11HyperVolume11aioInternalEP10DiskIO_IR PP8 ee2d70: 0xa55cd7 _ZN11HyperVolume11aioInternalEP10DiskIO_IR PP8 ee2e40: 0x133ccb5 _ZN10SnappedVol11aioInternalEP10DiskIO_IRP P8 ee2e70: 0xa55cd7 _ZN11HyperVolume11aioInternalEP10DiskIO_IR PP8 This occurred was because of a limitation in the fibre driver code to support up to 64

Code now breaks larger requests into chunks of 64 entries to prevent this sort of panic.

379947 / 36247340

311

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

request blocks only.

File System

The following entries describe fixed bugs pertaining to File Systems:

Severity

Symptom Description

Fix Summary

Tracking Number

1 If a frozen file system was unmounted while the system was out of vnodes, the naming node became inconsistent which led to a Data Mover panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 6d8dd7 Err code: 0 Target addr: 1d4 **. The stack trace showed:

0x2acb734: 0x13fb94 waitForReboot+0x90 0x2acb754: 0x13fda9 fault_dump+0x67 0x2acb774: 0x138d94 handle_fatal_fault+0xa 0x2acb7b0: 0x7dc2c4 _Z17MPDShadowCallBackP15File_NamingNodeP K18File 0x24a45b00: 0x7cf46a _Z14ShadowCallBackP15File_NamingNodePK18 File_N 0x24a45c00: 0x7777a6 _ZN9UFS_Vnode19doDirChangeCallBackEPK18F ile_Na 0x24a45c30: 0x694d7f _ZN12File_DirNode21lookupComponentLocked EPK18F 0x24a45d00: 0x69b627 _ZN16File_RootDirNode21lookupComponentLo ckedEP

The naming node is now appropriately replaced under this condition to avoid the inconsistency.

165384 / 33310518

312

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 When user tried to mount a root file system in read-only mode on a stand-by Data Mover, the Data Mover panicked with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 794170 Err code: 0 Target addr: 40 . The stack trace showed:

0x2898554: 0x13fa6c waitForReboot+0x90 0x2898574: 0x13fc69 fault_dump+0x67 0x2898594: 0x138cfc handle_fatal_fault+0xa 0x28985e0: 0x75723c _ZN9UFS_Vnode10lookupCompEPK18File_Name Componen 0xc48a77d0: 0xec4db4 _ZN10MGFS_Vnode10lookupCompEPK18File_Na meCompo 0xc48a7810: 0x6f7de3 _ZN18UFS_FileSystemKind6initFSEPKcP9IO_ Objecti 0xc48a7900: 0xec1f89 _ZN19DHSM_FileSystemKind6initFSEPKciR6O ptionP1 0xc48a7980: 0x6cc267 _Z25File_MountCommandInternalR6OptionP1 8VFS_Fi

During a mount, code now checks whether the default log device is initialized or not. If not initialized, the following error is returned: Log device not initialized.

370855 / 35172362

1 Issuing a quota check command at the same time as a quota operation from a Windows client resulted in a deadlock condition and rendered the file system inaccessible until the Data Mover was rebooted.

The fix has the quota command from Windows abort requiring the quota lock while the FS is inactive.

381651 / 36551080

1 1. During any version 5.6 to 6.0.36.4 upgrade, if the quota policy was set to filesize (which is not the default blocks configuration), the task for upgrading Data Mover could timeout because the Data Mover takes too long to reboot due to an automatic tree quota check. 2. While running version 6.0.36.4,

Code no longer forces a quota check during the reboot procedure, following a quota policy change. See Primus emc254145 for more details.

388085 / 37305372

313

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

changing the quota policy and rebooting the Data Mover could cause a long reboot, due to the automatic tree quota check.

FileMover (DHSM)

The following entries describe FileMover-related fixed bugs:

Severity

Symptom Description

Fix Summary

Tracking Number

1 User saw incorrect data when reading stub files on loopback connections. This issue occurred because of the way that offsets and read sizes were calculated on read operations which led to read timeouts.

Code now correctly adjusts the read offset and size on error recovery read operations.

378918 / 36077544

1 A Data Mover exhibited high CPU usage when DHSM was enabled and in-use on one or more of the file systems. Errors similar to the following were displayed in the server log. 1. "error direct read block xxx of entry xxx, status 20 - EndOfFile", and/or 2. "File size mismatch: Stub file xxxx bytes, migration vnode xxxx bytes,ino=xxxx status=EndOfFile"

Fixed a logic error in the DHSM I/O (read/recall) path to prevent this infinite loop.

372101 / 34673002

Install/Config/ Upgrades


Severity

Symptom Description

Fix Summary

Tracking Number

314

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 During an NS4-120 installation, the "setup_enclosure -detectSingleDMSystem" failed with the following error: Discover 1 Mgmtswitches in Primary subnet A, 0 Mgmtswitches in Secondary subnet B Please check cables and Mgmtswitches on Secondary subnet B Error: detectSingleDMSystem FAIL (EMISMATCHCOUNT) Remove Dynamic IP range in DHCPD_CFG ... OK Restart DHCPD with updated DHCPD_CFG ... OK Error: UNDEFINED retval = -10 (EMISMATCHCOUNT).

Code was modified so that the install will not fail if it does not discover the switch on subnet B in a single Data Mover system. It will be discovered by the setup_enclosure -initSystem command which is run later during the install process.

372086

2 While performing one of the following Upgrade/Conversion procedures, setup_slot failed while adding a new I/O module in single Data Mover system: -Upgrade from one blade to two blades -Add a second SFP for second Tape support -Add an I/O module (to all blades simultaneously) -Changing to the faster I/O Module of the same type -Add a second Control Station.

The wrong reason code was being monitored for these procedures, and because the system never reached that reason code (5), the command returned an error. This has been fixed.

374063

1 During Express install, the "set_up_Data_Mover" process took 40 minutes to complete on a 4 Blade setup. This was due to a built-in delay during a reboot.

The fix eliminates the delay when issuing a reboot with NAS services down.

388839

iSCSI

The following entries describe fixed iSCSI-related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

2 When client demoted an iSCSI snap LUN using the force flag, the iSCSI remove LUN operation logged an error instead of

The server log message for this case is not a warning. An

377716 / 33424708

315

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

a warning message in the server log. Users are likely to see this when using EMC Replication Manager because ERM internally demotes snap with force flag on its replication destination Celerra.

error message is only logged when the remove LUN operation actually fails.

NDMP

The following entries describe fixed issues related to Network Data Management Protocol and/or Backup:

Severity

Symptom Description

Fix Summary

Tracking Number

1 An uncooperative peer closed the TCP receive window during an NDMP backup. This led to the following panic because the NDMP was waiting on a TCP stream completion that never came. DART panic/fault message: >>PANIC: I/O not progressing LastVol touched Sn128 Kind 5 (ptr=0xa26d6104). The stack trace showed:

0x79dbe90: 0x13fb94 waitForReboot+0x90 0x79dbeb0: 0x13fda9 fault_dump+0x67 0x79dbed0: 0x13fc91 PANIC+0x29 0x79dbee0: 0x8478e0 _ZN9RawIO_IRP13checkProgressEv+0x11a 0x79dbfd0: 0x8477bc _ZN17IrpProgressThread5startEv+0x6 0x79dbfe0: 0x144344 Sthread_startThread_internal+0xe

This was fixed by setting the abort flag for 3-way VBB operation, should an unmount or internal error require abortion of the backup/ restore operation.

371282 / 35158192

Network

The following entries describe fixed network-related issues:

316

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 LACP could not be used with Dell Poweredge 6248 network switch and Celerra. The following error message appeared in the server_log: DRIVERS: 3: LAP_Port::receive_LACP len=114 expected_len=110. The verification of the PDU length was too strict, where it should have been verifying the PDU is a least the minimum size of 110.

The PDU length verification has been changed to accept anything not less than the expected PDU length.

371067 / 35186226

1 When the client sent a packet greater than the MTU size (between 1518 and 1530 bytes), the Broadcom device hung. The Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../dskdump.cxx at line: 1836 : network issue. The stack trace showed:

0x1f635ae0: 0x14124c waitForReboot+0x90 0x1f635b00: 0x1414d5 fault_dump+0x67

0x1f635b20: 0x1413bd PANIC+0x29 0x1f635b30: 0xa45118 _Z11paniccfgcmdR6Option+0x178 0x1f635c70: 0x1cbdab _ZN11cfgcmd_desc6invokeEPKcPP8CCMD_Msgii +0x1a7 0x1f635ce0: 0x2b205a _Z10xml_cfgcmdP7macDataRP8CCMD_Msg+0x392 0x1f635d30: 0x2ad00c _ZN17macCCMDCommand_DB6invokeEPKcP7macDa taRP8C 0x1f635d60: 0x2ad512 _Z17mac_CCMDDocParserP7macDataRP8CCMD_Ms g+0x20

Fixed the Broadcom 10G hang issue by correcting the MTU (max transmission unit) size passed to the device.

376090 / 35799674

1 An exploit of the NFS mount protocol was identified where an attacker could spoof IP addresses that are normally used between the Celerra Control Station and Data Mover over a private IP network and gain unauthorized access to file systems on the Celerra.

Code has been fixed to reject all mount requests using the Control Station IP addresses that do not come over from a private network that is internal to the Celerra

118547

317

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

cabinet, eliminating this vulnerability.

NFS


Severity

Symptom Description

Fix Summary

Tracking Number

1 When using NFSv4, a missing reference for NFSv4 state object on Data Mover caused a Data Mover to panic with the following DART panic/fault message: *** Page Fault Interrupt. Virt ADDRESS: 66fdf2 Err code: 2 Target addr: 24 **. The stack trace showed:

0x2c2aba4: 0x141254 waitForReboot+0x90 0x2c2abc4: 0x1414dd fault_dump+0x67 0x2c2abe4: 0x13a3a0 handle_fatal_fault+0xa 0x2c2ac60: 0x66e352 _ZN16NFSv4_StateTable5CloseEmP8stateid4P 8nfs4_c 0xe10b5d50: 0x65d129 _Z10nfs4_closeRP4msgbS1_RPhR8nfs4_ctx+0x 18b 0xe10b5d90: 0x6592b0 _Z23nfs4_convertNfsv3Statusm+0x626 0xe10b5ee0: 0x63b64d _ZN17nfs_serverInitial5startEv+0xad7 0xe10b5fe0: 0x145bd6 Sthread_startThread_internal+0xe

Code has been fixed to properly handle missing reference on the object and prevent this panic.

372545 / 35327318

318

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

2 NFSv4 locks could not be reclaimed after a temporary network outage. If an NFSv4 client temporarily lost contact with the Celerra and attempted to reclaim locks after its lease expired (20 seconds by default), the operation would fail. Access to the file from the process that owned the lock could be impacted, though if the process that owned the lock specifically released the lock, access would be restored. This problem was introduced in 5.6.45.

Fixed a miscomputation of the NFSv4 owner sequence number in Open-Claim-Previous operation. Attempts to reclaim locks will now succeed.

377431 / 35858230

1 NFSv4 code did not check permissions when modifying the ACL based on the mode bits. This can result in access behavior inconsistent with permissions.

Code now correctly applies permissions when changing the NFSv4 ACL.

378959 / 36091074

2 Unable to write data to a File Level Retention Enterprise (FLR-E) files system if the file system was mounted using NFS V4.

Code has been fixed to ensure users can write data to FLR-E file systems mounted using NFS V4.

368246

1 In an unhealthy network, where many network packets were lost, memory on the Data Mover was exhausted by those many dropped requests. This led to a Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../strbufs.cxx at line: 1747 : out of msgb. The stack trace showed:

0xe2d6960: 0x20adcc _ZN15Buffer_MsgbPool3getEv+0xac 0xe2d6990: 0x20af9a allocb+0x54 0xe2d69c0: 0x6e5ec8 nfs_rput+0x2c0 0xe2d6ab0: 0x40aa3e udp_input+0x9fa 0xe2d6c00: 0x443879 _Z9ip6_inputP10IPDevice_tP4msgbi+0x10cb

Code has been fixed to properly release allocated memory for dropped requests, preventing Data Mover panics in unhealthy network environments.

379200

319

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

0xe2d6d60: 0x4316f0 _Z10ip6_mxrputP5queueP4msgb+0x13a 0xe2d6dd0: 0x37b4bd _Z7bcmreadPvP4msgbtjj+0x4e7 0xe2d6ee0: 0x38b2b2 _Z20LM_ServiceInterruptsP16_LM_DEVICE_BL OCK+0x3

Replication

The following entries describe fixed problems encountered while using Celerra Replicator, SnapSure or Checkpoints:

Severity

Symptom Description

Fix Summary

Tracking Number

1 The Data Mover on the destination side of a ReplicationV2 session panicked after a failover issued using the session name. This happened because the replication database contained an old entry for an inactive session with the same name and the failover operation used the old entry rather than the new one.

The session ID is now used during the failover since it will always be unique.

360751 / 34233340

2 When attempting to create an iSCSI replication session, if version set was out of space, the task showed success when it actually failed.

The iSCSI replication create task status was changed to Severity=error. The following will be displayed in the task status: Version set out of space.

374003

320

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 In a cascade replication scenario, a replication session for the middle hop to the third hop was created before creating a session for the first hop to the middle hop. The code allowed a file system that acted as replication source and destination (middle hop) to be mounted as read-write. This violated replication destination read-only property assumption and caused a corruption on that file system which lead to Data Mover panic with the following DART panic/fault message: >>PANIC in file: ../ufsdirhash.hxx at line: 623 : UFS_DirHashIter_dir3::getEntry: mangled entry. The stack trace showed: 0x2b1f7050: 0x140ce4 waitForReboot+0x90 0x2b1f7070: 0x140f69 fault_dump+0x67 0x2b1f7090: 0x140e51 PANIC+0x29 0x2b1f70a0: 0x7b045a _ZN20UFS_DirHashIter_dir38getEntryEm14Bu fInten 0x2b1f70e0: 0x7c2e9d _ZN20UFS_DirHashIter_dir34findEmPKcm13Fi le_Nam 0x2b1f7110: 0x7d1dd4 _ZN20UFS_DirHashIter_dir34findEmPK18File _NameC 0x2b1f7150: 0x7c3ad7 _ZN15UFS_DirOps_dir37scandirEP9UFS_Vnode PK18Fi 0x2b1f71d0: 0x7cd806 _ZN15UFS_DirOps_dir311getSynonymsEP9UFS_ VnodeP

A new check was added to the code, which prevents the read-write mount of a file system that acts as replication source and destination in a cascade replication.

366022 / 34665376

2 While trying to perform an out-of-family Rep V2 replication, whenever a server with a lower slot number was made as standby for a server with a higher slot number, the multicast session lost the lower slot number server from the target list. When user invoked the nas_fs -delete command, the following error was generated: Error 5005: DpRequest_Max_VS_SuspendedCheckpointsReach ed.

When there is a change in the server state, CIC (Celerra Interconnect) multicast target list is now generated from scratch rather than being modified. This solves the problem of a target being missed due to state and relation change of the servers.

144857

321

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

2 While using Replication V2, if a Data Mover was rebooted during a nas_copy process, the data transfer completed successfully, but the command did not complete and hung after the reboot.

The nas_copy command will now complete after a Data Mover reboot.

376684

1 During merge operations involving file system checkpoints, the TOC (Table Of Contents) became inconsistent with the trailer and map. This inconsistency led to the following panic which required an unmount of the bad checkpoint before the Data Mover could be stabilized. DART panic/fault message: D130832_0: Btree_Insert failed, mark ckpt corrupted. The stack trace contained various messages similar to the following:

0x20643f40: 0x141254 waitForReboot+0x90 0x20643f60: 0x1414dd fault_dump+0x67 0x20643f80: 0x1413c5 PANIC+0x29 0x20643f90: 0xfb36ad _ZN28IndexChainCorruptMarkRequest13handl eReque 0x20643fc0: 0xfb3312 _ZN23CkptCorruptMarkerThread5startEv+0x80 0x20643fe0: 0x145bd6 Sthread_startThread_internal+0xe

Code now verifies the TOC entries against the other 2 references before the merge, during the merge, and after merge completes. Code then corrects the TOC as needed.

366985 / 34775056

1 Network issues might cause RepV2 session hang.

Code now promotes the RCP_CLOSE message priority so it will not cause blocking.

156084

322

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 If the internal checkpoint name exceeded 16 characters, memory corruption occurred which caused a system panic such as I/O not progressing and volume display error. DART panic/fault message: >>PANIC in file: ../sched.cxx at line: 846 : SYSTEM WATCHDOG. The stack trace showed: 0x5089f24: 0x14124c waitForReboot+0x90 0x5089f44: 0x1414d5 fault_dump+0x67 0x5089f64: 0x1413bd PANIC+0x29 0x5089f74: 0x1432c9 system_watchdog+0x77 0x5089f84: 0x139f6a _t_nmi+0x68 0x5089fe0: 0x145bfa Sthread_startThread_internal+0xe

Checkpoint internal names whose length exceeded 16 characters were truncated resulting in unexpected behavior - for example the name D2354619762_10113 could be truncated to D2354619762_1011 which does not exist. The naming convention was changed to shorten the name. The new format of the name is "D<ckptFsID>_<ckptGenCnt>", which keeps the total length below 16 characters and maintains the uniqueness of the name. This checkpoint name is only used internally for debug/display purposes. It can be seen in the server_log during checkpoint refresh operations for example - 2010-09-20 06:42:01: 26042826752: SVFS: 6: D2354619762_1011: Recovered(Stable Index) FsID:23546 state:Active SavVol:29390 FsVol:D2480619762_0 rsv=2 h=12322040 c=12322057 t=12322057, in 0 Msec 2010-09-20 06:42:01: 26042826752: SVFS: 6: D2354619762_1011:

383673 / 36791874

323

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

currDel:D2354619762_101 13 st:Active prevDelta:1573888

1 Replication V2 sessions were not automatically synchronized. The timer being used to check if it is time to transfer was using an overflowed value.

The interval between checks of whether it is time to sync a session now uses a consistent value rather than a calculation which could overflow.

386995 / 37192564

Security

The following entries describe fixed security- related issues:

Severity

Symptom Description

Fix Summary

Tracking Number

324

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 The LDAP configuration on a Data Mover was using multiple LDAP servers, and LDAP was configured to use the Active Directory "member" attribute instead of "memberUid". When the network connectivity to all LDAP servers was bad, memory corruption occurred and the Data Mover panicked with the following DART panic/fault message: >>PANIC in file: ../malloc.cxx at line: 351 : Memory corrupted . The stack trace showed:

0xdb9477e0: 0x14124c waitForReboot+0x90 0xdb947800: 0x1414d5 fault_dump+0x67 0xdb947820: 0x1413bd PANIC+0x29 0xdb947830: 0x1fe8e7 free+0x69 0xdb947860: 0x1fefb8 _ZdlPv+0x10 0xdb947870: 0x4f6582 _ZN10LdapClientD1Ev+0x32 0xdb947890: 0x53f311 _ZN16LdapDomainClient13getgrbymemberEPKc PPP12A 0xdb947bd0: 0x69b7f3 _ZN20Access_GroupDataBase15addNewExtraGi dsEP18

Code was modified to correctly handle failover between LDAP servers and prevent memory corruption.

376153 / 35803348

1 NTP software on the Control Station was vulnerable to the denial-of-service attack described in US CERT VU568372 when NTP service is running (disabled by default).

Code has been fixed to pickup the non-vulnerable version of NTP software from RedHat which prevents this denial-of-service attack.

364874 / 34476786

Storage

The following entries describe fixed storage-related issues.

325

Version 6.0.40.8


Severity

Symptom Description

Fix Summary

Tracking Number

1 A Data Mover watchdog panicked due to FC environmental conditions. This can happen when the processing used to clear some FC conditions takes too long due to unnecessary reads across the PCI bus.

The unnecessary reads on the PCI bus have been removed during FC processing.

369324 / 35030006

1 Direct attached tape drives connected with cable lengths greater than 50m may not be discovered after upgrade to 5.6.47.11 and above. The issue was caused by a race condition that was triggered by increased latencies using longer cables.

The code has been fixed to avoid race condition when latency is increased due to long cables.

381342 / 36335402

UFS


Severity

Symptom Description

Fix Summary

Tracking Number

1 In an environment with VMware ESX server that was using Celerra iSCSI LUNs for its VMFS file systems, the ESX server had aborted an iSCSI session and disconnected from iSCSI LUNs while a large iSCSI snap was being deleted on the Celerra. This caused an outage on the ESX virtual machines. The following messages in server_log indicate the issue: SCT: 6: Task Management Function received: Abort Task, I_T Nexus: [<iqn>] ISCSI: 3: IscsiSession::closeSession: I_T nexus lost, sess=0xe0b6dc04 from [<iqn>]

The Data Mover code has been updated to improve response time during the iSCSI snap delete process. From the ESX server side, VMware updated ESX server code to allow longer timeout before aborting an iSCSI session.

361139 / 34143216

326

Troubleshooting and getting help


Troubleshooting and getting help

This section provides specific instructions for contacting EMC Customer Support and obtaining additional information about EMC products.

Where to get help

EMC support, product, and licensing information can be obtained as follows.

Product information: For documentation, release notes, software updates, or for information about EMC products, licensing, and service, go to EMC Online Support (registration required) at:

http://Support.EMC.com

Troubleshooting: Go to EMC Online Support. After logging in, locate the appropriate Support by Product page.

Technical support: For technical support and service requests, go to EMC Customer Service on EMC Online Support. After logging in, locate the appropriate Support by Product page and choose either Live Chat or Create a service request. To open a service request through EMC Online Support, you must have a valid support agreement. Contact your EMC Sales Representative for details about obtaining a valid support agreement or to answer any questions about your account.

Note: Do not request a specific support representative unless one has been assigned to your particular system problem.


327


Copyright © 2013 EMC Corporation. All Rights Reserved. Published in the USA.

Published February 2013

EMC believes the information in this publication is accurate as of its publication date. The information is subject to change without notice.

THE INFORMATION IN THIS PUBLICATION IS PROVIDED "AS IS." EMC CORPORATION MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO THE INFORMATION IN THIS PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Use, copying, and distribution of any EMC software described in this publication requires an applicable software license.

EMC2, EMC, and the EMC logo are registered trademarks or trademarks of EMC Corporation in the United States and other countries. All other trademarks used herein are the property of their respective owners.

For the most up-to-date regulatory document for your product line, go to the technical documentation and advisories section on the EMC online support website.