In Docker Swarm stack deploy, gitlab_config owner changing from nobody to root

Summary

We are trying to perform a DR test which requires our storage team to export a clone of production and mount to the DR environment at the time of the test. We unfortunately do not have active/active failover.

We use Omnibus in a Docker Swarm environment. All of the production volumes are owned by nobody:nobody with 775 perm. When they clone and mount the prod export in DR, we notice that the gitlab config volume gets changed to root. We change the owner back to nobody but then when we run the stack deploy, it once again changes back to root and fails saying there was an error mounting the volume. All other volumes seem to have no issues but we can't start the gitlab app. So we are not sure what happens during the clone/mount that would cause the owner of config to change to root and thus cause the containers to fail to mount.

Steps to reproduce

  1. Export production volumes
  2. Mount to DR mount point.
  3. Notice gitlab_config volume change owner from nobody to root.
  4. Change volume owner back to nobody with 775.
  5. Run stack deploy.
  6. Notice gitlab_config volume once again changes owner back to root and Docker shows error in mounting volume.

Relevant logs and/or screenshots

  • Before running stack deploy

gitlab1

  • Error mounting volume

gitlab2

  • Owner changed back to root

gitlab3