From 1,000 to 5,000 Users: Scaling GitLab Without (Too Much) Pain
summary
A GitLab operations team at Michelin explains how they scaled GitLab from 1,000 to 5,000 users by expanding from 1 VM to a 28-node deployment and reworking components like Rails, Gitaly Cluster, Praefect, Sidekiq and Puma. They describe handling painful deprecations (for example moving off NFS), redesigning backup and restore strategies, paying down technical debt, and which technical choices helped or hindered them, with concrete practices for Ops/SRE teams to keep a large, evolving platform under control.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:30Michelin IT ecosystem
- 9:53Initial GitLab deployment
- 15:31Manual scaling attempts
- 21:49NFS failure and fix
- 27:12GitLab Environment Toolkit demo
- 35:49Scaling best practices
- 40:25Q&A
related talks
29mScaling Across the Enterprise
38mHow We Scaled GitLab For a 30k-employee Company
27mB How GitLab scaled Git access with Go and what we gained from it
21mDriving Cultural Change to Improve Software Development and CI/CD
37mPanel: Building a GitOps...
20mBuilding Real-Time DevOps with Open Source
29mBest Practices for Kubernetes Runners
26mRunning GitLab and GitLab CI in Kubernetes for the Enterprise