From 1,000 to 5,000 users: scaling GitLab without (too much) pain.
REXCloudInfrastructureOpen Source
2026-12-09 | 12:00 AM - 12:25 AM
Information
Going from 1,000 to 5,000 users, dealing with the sudden deprecation of legacy architectures, completely rethinking our backup strategy, and absorbing growing technical debt while keeping a critical service running…
In recent years, it has often felt like we were in a race against our own infrastructure. Several times, we thought the
beast was completely getting away from us. And yet, in 2025, we finally took back control!
In this 100% hands-on talk, I’ll share our experience running GitLab at scale in a large-scale IT environment:
- The real scaling pitfalls (from 1 VM to 28 nodes, Rails, Gitaly Cluster, Praefect, Sidekiq, Puma, workload distribution, etc.)
- How to survive the deprecation of legacy architectures (e.g. moving away from NFS) and rebuild a reliable backup/restore strategy
- The technical choices that saved us (and the ones we regret)
- Best practices and lessons learned to stay in control of a constantly evolving platform
- How to survive the deprecation of legacy architectures (e.g. moving away from NFS) and rebuild a reliable backup/restore strategy
- The technical choices that saved us (and the ones we regret)
- Best practices and lessons learned to stay in control of a constantly evolving platform
Ops, tech leads, SREs, and tech enthusiasts: come see how we tamed the beast—and leave with solutions you can put into practice right away!
Speakers
-
Adrien GOORIS
GitLab sans (trop) souffrirDéveloppeur -
Lucas DRAGO
MichelinExpert technique