Most teams back up GitLab and their registries. Far fewer have restored them. We take your real backups, restore them into an isolated environment, measure how long it takes and list everything that was missing — before you find out on a bad day.
Backups run every night. Nobody has tried a full restore.
Restore GitLab, registries, secrets and storage into an isolated environment.
Real restore times, a gap list and a written runbook.
From GitLab's own documentation:1
We agree the list on day one. Typical scope:
| System | What we restore and check |
|---|---|
| GitLab | Database and repositories, plus configuration, secrets file and object storage; log in, open projects, run a pipeline |
| Registries | Harbor, Nexus or GitLab registries: pull a sample of images and packages after restore |
| Secrets store | Vault or OpenBao snapshot restored and unsealed; read a test secret |
| Object storage | Bucket contents and versioning; sample objects compared with the source |
| IaC state | State files restored and a plan run against them |
Worked, partly worked or failed — with evidence.
How long each restore really took, compared with what you need.
Missing files, keys, steps and people, ranked by impact.
Step-by-step instructions written during the test, ready for the next time.
Shown for a typical mid-size setup. Larger platforms take longer; the scope and price are fixed on the first call.
Agree systems, get access to backups, prepare an isolated environment.
Restore each system, test it, time it, write down every step.
Results, gaps, runbook and a short walk-through.
Fixed price, agreed on the first call.
No. We read backups and restore them somewhere else. Production keeps running as usual.
Then the test did its job. You get the exact reason and the fix, while there is still time to apply it.
Yes. A restore test every quarter is part of Open-stack care.
If you are not sure, it is time. A 30-minute call is enough to scope the test.