SVCPLAN-8877: Add logic to return codes in case of backup failures

[inwho]

In cases where backup jobs terminate before completion (runtime error/timeout etc), the resultant files (/var/log/backup_last_run and /var/log/backup_check_last_run) do not report a job status code, which then lead to empty fields in $backup_return_code, $prune_return_code, and $return_code.

This change checks to see if the variables are empty (which implies a failed job run) and if they are, has it report an error code 1.

Consideration: maybe it should return a different error code to signal that the backup job script itself failed?

Script tested locally on cc-adm01.

[billglick]

While your changes fix the issue with incorrect return codes, I still think we need to also figure out why the return codes are failing.

[inwho]

I've been digging into it and unfortunately can't find a common factor across clusters yet.
Currently working on mg-adm01 which seems to relate to borg backup failures.

cc-adm01 is a bit odd since it sometimes works sometimes doesn't. I wonder if it is related to the recent instability issues.

[billglick]

This looks good.

I also could not figure out why it was sometimes failing.