backup should not be aborted on error

[nielsole]

I am currently evaluating ark and tested the backup & restore.

Apparently I ran into some race condition so that a PV was already deleted.

time="2018-01-10T15:30:20Z" level=error msg="backup failed" error="[persistentvolumes \"pvc-<redacted>\" not found, persistentvolumes \"pvc-<redacted>\" not found, persistentvolumes \"pvc-<redacted>\" not found, persistentvolumes \"pvc-<redacted>\" not found]" key=heptio-ark/initial-backup2 logSource="pkg/controller/backup_controller.go:258"

The backup was subsequently aborted (I am not sure whether this was the reason, just a correlation). Wouldn't it be better to conclude the backup as far as possible and still show the backup as Error or Partial?
Better to have something than to have nothing. Especially when the bigger the cluster gets, the more likely it is that there is some transient error.

[ncdc]

Hi @nielsole, thanks for your comments! This is similar to #225 (comment). I think we'll probably want to make backups similar to restores, where we try to do as much as possible, recording warnings & errors along the way.

[nielsole]

hadn't seen that one. Thanks for the quick reply