Proxmox Backup Server full

[jmvf2004]

Hi, I've been recreating a bug in PBS. When it gets full, I need to manually delete chunks and erase datastore in order to free space, because when I delete some backups no space seems to be released. Garbage collection won't help either.
I'm now testing with SDA and SDB instead of RAID 0, same problem. Other machine on production, same problem.
Any help?

 

[shanreich]

Garbage Collection will only remove unused chunks that haven't been used for at least 24 hours (24h5m to be exact). If this is not the case, garbage collection will not remove the chunks. Chunks that are within that grace period get logged as Pending Removals at the end of the log output of the garbage collection task.

 

[VictorSTS]

I mostly use ZFS, which as any COW filesystem, if disk gets too full you can't even delete anything. I do always leave a 10-20 gigabyte file on every filesystem holding a datastore that I can delete in case something goes wrong and the filesystem gets to 100%.

 

[marlon006]

I mostly use ZFS, which as any COW filesystem, if disk gets too full you can't even delete anything. I do always leave a 10-20 gigabyte file on every filesystem holding a datastore that I can delete in case something goes wrong and the filesystem gets to 100%.

I can't believe i have to do this in the future to stop pbs from breaking itself. And filling the filesystem is super easy, since pve doesn't do any freespace check, it just dumps what it can and suddenly cant seem to grasp why it can't continue backing up. The ability of setting quotas would fix this but instead this issue is being completely ignored.

 

[David Thistlethwaite]

Try this:

1. Your current backups dedup ratio's are telling what kind of dedup efficiency your are getting with your current data set(s)
2. plan for enough disk space, if you don't have enough disk, no backup system will work for you
3. review your backup retention parameters (prune parameters).
4. set a quota on the underlying file-systems containing the PBS datastore(s), leave 20% overhead as this works for most file systems
5. before the next backup, prune
6. 24 h 10 m after the prune job run GC on the datastore this will cleanup what can be cleaned up
7. if you still have space issues see points 2 and 3