borg delete/compact randomly does not return

[zoot]

I have a script which lists all archives to find the oldest one to delete.

The repo is quite large, with almost daily backups back to 2024.

Once the oldest archive $arch is found:

borg delete --debug --show-rc $user@$host:$repo::$arch

... and then if it returns 0, then:

borg compact --debug --show-rc $user@$host:$repo

It often fails to return / exit, hanging indefinitely.

I suspect it's at the compact step, but will update this thread when it occurs again.

I thought it may relate to SSH timeouts, but adding the following has not remedied the behaviour:

$ cat .ssh/config
Host *.backup.host.provider.domain
ServerAliveInterval 30
TCPKeepAlive yes

I sheepishly admit that this host is running borg 1.2.7 which may mean the issue is already resolved in a later release.

[ThomasWaldmann]

Unclear:

• how long did you wait for borg compact?
• borg compact can take a long time, depending on how many segment files are actually compacted.
• yes, could be also the ssh connection getting killed.

[zoot]

Thanks for your very prompt response @ThomasWaldmann

The issue is that sometimes the compaction step completes quickly, but other times appears to hang indefinitely (hours).

I have been much more patient with repo check, which can take hours / overnight.

For the next invocation, I will background the process and write to a log and time the process for more clarity.

I may have to rewrite the script to handle SSH process death/hanging/timeouts better.

[ThomasWaldmann]

Guess in the worst case (having to compact all segments), borg compact could take similarly long as borg check.