Documentation update: Garbage-collect for container registry in HA environment
Background:
sudo gitlab-ctl registry-garbage-collect
will shut down the registry and run the Docker garbage-collect
feature to remove any unreferenced manifests and reclaim disk space for their registry.
Unfortunately one of our customers ran this command on an instance in an HA environment while one of their other instances was still writing to the registry storage which removed referenced manifests and resulting in a long, manual restoration process.
My proposal here is to update the documentation warning at https://docs.gitlab.com/omnibus/maintenance/README.html#container-registry-garbage-collection from:
- Warning The command below will cause Container registry downtime.
to
- Warning The command below will cause Container registry downtime. If running in an HA environment please stop all active container registries or put into read-only before proceeding.
I am opening this issue to discuss whether this is the best solution, or whether we would rather implement a new feature to the app that would shutdown all existing registries or put into read-only mode via the API.