Proxmox and PBS retention keeps the newest backup in each calendar bucket, not a span of days. How a month of failed jobs can prune away your history, how the keep-* rules combine, and how to set retention in one place.
Free weekly email with the week’s posts, videos and one tip: https://chartmann1590.github.io/homelab-the-right-way/newsletter/
Disclosure: this is a Homelab the Right Way article by Charles Hartmann, written with AI help and edited by him; the setup described is his own.
so a retention policy that looks like three months of history on paper can collapse to a single backup, with no error anywhere, if the jobs stop running.
How? If you have keep-monthly=3, it would still keep at least three? Having daily, weekly and monthly is a different thing.
Great question — it’s about calendar buckets vs absolute counts. With keep-daily=7 you get at most 7 daily files (one per day). But keep-weekly=4 means the newest backup of each of 4 weeks, not necessarily 4 weeks of backups if some weeks are missing. If no jobs ran for a few weeks, PBS has nothing new to put in those weekly buckets, so they stay empty. The monthly bucket acts the same way — it only keeps the newest backup per month that actually had one. So when jobs stop entirely, you don’t lose the last daily (it’s within keep-daily=7), but everything older than 7 days gets pruned because there are no newer files to anchor into the weekly/monthly buckets.
That doesn’t explain how “3 months of history collapse into a single file” at all


