Click here to load reader
Upload
alexanderwerner
View
338
Download
0
Embed Size (px)
DESCRIPTION
Most free software projects happen in a virtual environment, and build on a joint cooperation of many contributors worldwide, breaking barriers of language, culture and time zones - something that would have been unthinkable with the technology available 20 years ago. Essential to the success of any such free software project is a reliable and scalable infrastructure that grows with the community and scales based on its needs. Being one of the largest free software projects worldwide, the LibreOffice community was challenged with providing a professional infrastructure from the very first day. Following our principles, nearly all components solely rely on free and open source software. We not only build on third party components, but also develop many of our infrastructure tools on our own, and make them available to the worldwide community, while at the same time contributing to projects we make use of. This talk gives an overview on the infrastructure of the LibreOffice project and how it provides the playground the community and the software need to grow and flourish.
Citation preview
1
The infrastructure ofThe infrastructure ofthe LibreOffce projectthe LibreOffce project
Alexander WernerThe Document FoundationLibreOffce Conference Bern4 September 2014
2
About Alex
A long long time of activityfor free softwareMember of The Document Foundationresponsible for the project's infrastructure as freelancerPython enthusiast
3
Looking into the engine room
4
Controlled chaos
Own dedicated servers
Dedicated servers rented by others
Hosted services
Nginx
Apache
MySQL
PostgreSQL
Deployment on bare metal
VMs
UFW
Shorewall
Postfx
Mirrorbrain
Askbot
Plone
Silverstripe
Debian 7
Ubuntu 12.04
Amavisd
Mailman
Mlmmj
Saltstack
Owncloud
Mediawiki
Dovecot
Mirmon
Etherpad
Planet
Etc. pp.
5
Jump to the next level
6
Taming the chaos
Goals
High availability of services
Fewer services with the same purpose
Better maintainability
Better use of resources
Easier scalability
7
Taming the chaos
Reaching the goals
High availability
Use current virtualization and storage technologies
Build a cloud-like infrastructure
Get better suited hardware
Fewer services with the same purpose
Find duplicate services
Choose the one that works best
Better maintainability
Reduce the number of different speced servers
Reduce confguration complexity
8
Taming the chaos
Reaching the goals
Better use of resources
Software that needs less CPU/RAM
Looking for simpler confguration
Easier scalability
Get hardware that is suffcient for quite a number of VMs
Make use of an infrastructure/cloud provider
9
Our new engine
64 cores256GB RAM4x2TB Enterprise Level SATA HDDsHardware RAIDRedundant power supplyDirect LAN connection between two of themIPMIenough IPs (both v4 and v6)
10
The gearbox
oVirt as infrastructure provider
Provides virtual machines, virtual storage and virtual networkson top of CentOS 6.5Running in self hosted engine mode – management engine as VMGlusterFS as replicated storageMigration of VMs between hosts without downtimeKVM as virtualization technology
11
Our chauffeur
SaltStack for infrastructure orchestration
One master commands multiple minionsSimple, yaml based confgurationVery fast and parallelBootstrapping of the whole platform possibleSetup of new VM (production ready) from scratch within 20 minutesDocumentation of the confgurationMost parts open sourced (github.com/tdf/salt-states-base)
12
The whole platform
Easy to give new admins fne-grained access due to salt-based user management
Easy to maintain: changes to the salt-master propagate to the whole infrastructure
Highly available: VMs failover to the other host
Effcient use of resources: VMs can scale very large, memory overcommit included
It scales easily: VMs and additional hardware hosts can be easily provisioned
13
The masterplan
14
What's next
Writing salt-states that provision the platform on testing hardware
Allocation of the hardware by the hoster
Provision the nodes
Move the existing VMs to the new platform, starting with the staging ones
Move services from exisiting bare-metal machines to separated Vms
Change from Apache to Nginx, from MySQL to PGSQL etc.
15
Join us, it's fun!
Infra confcall on Tuesday, September 9, 17:00 CET
Help appreciated in all areas
Virtualization
Backup
Monitoring
SSO
LDAP
Databases
Webserver
Mailserver
16
Communication channels
#tdf-infra at [email protected] mailing listsubscribe with [email protected]
own wiki category at https://wiki.documentfoundation.org/Infra
mail the admin team at [email protected]
All text and image content in this document, unless otherwise specifed, is licensed under the Creative Commons Attribution-Share Alike 3.0 License
This does not include the LibreOffce name, logo, or icon
17
Thank you for your attention!
Any questions?E-Mail: [email protected]: awerner