[Pull mirroring] LFS objects fetch timeouts

Problem

The pull mirroring process includes several steps:

1. Fetch remote changes
2. Update tags
3. Update branches
4. Update LFS objects

First three steps raise an UpdateError when they encounter a problem. That stops the pull mirroring process and notifies users about the error.

However, "Update LFS objects" step doesn't do that. We only log an error and mark pull mirroring job as successful. See errors examples in Kibana

UpdateError exception code was commented out because of gitlab-foss#61834 (moved).

But #414003 (closed) highlighted that we also have timeout errors coming from ListLFSPointers GRPC. We have a timeout setting equal to 30 seconds by default.

Questions

• What happens to mirrored repositories that failed to fetch LFS objects?
• How should we fetch LFS objects if it takes longer than 30 seconds?
• How should we notify users about LFS fetch errors?