zalando / spilo

Highly available elephant herd: HA PostgreSQL cluster using Docker
Apache License 2.0
1.57k stars 389 forks source link

Backups not being deleted as expected #937

Open a-thomas-22 opened 1 year ago

a-thomas-22 commented 1 year ago

Script does not delete old backups due to the LEFT check

Description:

When running the backup script on a weekly basis (via cron schedule 0 0 * * 0), the older backups are not getting deleted as expected. The core issue lies in the LEFT check in combination with the BACKUP_NUM_TO_RETAIN value.

Steps to Reproduce:

  1. Set BACKUP_NUM_TO_RETAIN to 4.
  2. Run the script weekly.
  3. Observe that after 5 weeks, there are 5 backups and none of the older backups have been deleted.

Expected Behavior:

After the number of backups surpasses the BACKUP_NUM_TO_RETAIN value, the oldest backups that exceed this count and are older than the DAYS_TO_RETAIN threshold should be deleted.

Actual Behavior:

Older backups are never deleted, regardless of their age, due to the LEFT check. Since we're only creating 1 backup per week and DAYS_TO_RETAIN is set to 4 days, the condition for deletion is never met.

creckord commented 11 months ago

See also issue #425. Not sure if the suggested change over there still applies after the changes from #938.

a-thomas-22 commented 5 months ago

I made a sidecar workaround for kubernetes until a PR to fix is made for those of us having this issue https://github.com/a-thomas-22/wal-g-pruner