As discussed at the link below, if Discourse is configured to upload backups to cloud storage, the upload can fail silently, leaving the tar file on local disk. After a few such failures the forum goes down because there is no disk space. Request a notification when an upload fails, like there is when a backup fails.
I think I have encountered this issue a couple of times on my forum as well.
The backup process is supposed to clean up when it’s finished, even when there was a failure. But it’s possible that the OS kills the backup process before it’s able to do that cleanup. Reasons I can think of right now:
- the host restarts (maybe because of OS upgrade)
- the Docker container restarts (maybe because the Docker daemon is updated or crashed)
- the system runs out of memory and kills processes
- the process crashes for whatever reason and isn’t able to exit cleanly
I guess the feature request here is: A background job should check if a backup was aborted. In that case it should cleanup temporary artifacts of the backup and send a warning to the admin.