What's new with CentOS at Facebook
summary
Facebook runs millions of machines and uses CentOS on 100% of servers and top‑of‑rack switches, with an ongoing migration to CentOS Stream on about 3,000+ boxes so far. Package tooling uses rpm/dnf but the rpm DB backend is rolled to NDB fleet‑wide, yielding zero instances of corruption; a daily pipeline pulls systemd master at 10:00, builds and smokes tests. An internal kernel (eBPF, cgroups, PSI) ships quickly via automated kernel updates, chef provides layered attribute config, and 1% canary rollouts proceed over two weeks.
AI-written summary · about
chapters (13)
- 0:00Introduction to infrastructure
- 1:14Operating system philosophy
- 3:40CentOS adoption and workflow
- 5:50Kernel development and management
- 7:35Systemd and package tooling
- 9:18Network management strategies
- 11:17Packaging and RPM database
- 13:48Provisioning and containers
- 16:05Configuration with Chef
- 18:57Major OS upgrades
- 21:36Rolling updates and testing
- 23:30Migration to CentOS Stream
- 24:54Q&A and final thoughts
related talks
36mCentOS Stream at Facebook
50mUpgrading CentOS on the Facebook fleet
35mBuilding the Community Enterprise Operating System through CentOS Stream
15mrebuilddb at Facebook scale
46mCentOS Stream: What you need to know
25mCentOS Stream: Progress so far
36mLinux Kernel Building, Testing, and Deployment at Facebook
28mLinux distributions as an ultimate challenge of continuous integration and how we solve it.