Feature request: snapshot retention

[rauanmayemir]

I'm setting up dragonfly instance with a following snapshot config:

spec:
  # ...
  snapshot:
    cron: "*/2 * * * *"
    persistentVolumeClaimSpec:
      accessModes:
        - ReadWriteOnce
      resources:
        requests:
          storage: 20Gi

It works beautifully, however, when left unattended, instance can rack up thousands of backup files within a month.

What would be great is if operator cleaned up older backups, something like:

spec:
  # ...
  snapshot:
    cron: "*/2 * * * *"
    persistentVolumeClaimSpec:
      accessModes:
        - ReadWriteOnce
      resources:
        requests:
          storage: 20Gi
    retention: "1d" # only keep up to 1 day worth of backups

[avdhoot]

It can be implemented on the operator side, but it’s preferred on the upstream. I think.. dragonflydb/dragonfly#2824

[Abhra303]

@rauanmayemir feel free to contribute. Can you share the design idea to support this feature? To me,the retention only makes sense if you configure volumes for snapshots.

[rauanmayemir]

@Abhra303 My initial thought for this is to have an operator sidecar that periodically runs a command that clears all snapshot files older than 'now - retention duration'. What constitutes older is the question, timestamp suffixes from dump filenames seem like a good way to detect that.

(But maybe introducing a sidecar for this is too much)