Limits gitlab-backup repository backup so that it can be used and tested on GitLab.com

Problem to solve

Backing up many terabytes of repository data for an extremely large GitLab could take days. With incremental and concurrent backups directly to object storage, incremental backups should be much faster, but taking an initial backup will take a very long time.

We need to add sufficient limits and controls to make repository backups usable at GitLab.com scale.

Further details

Proposal

• Limit gitlab-backup create by storage, so that it can be rolled out progressively
• Limit gitlab-backup create by number of repositories bundles that may be generated