Hi,

I have a three node PVE cluster running with a few VMs.

One of them, I back up every two hours.

And every time the backup runs (and completes successfully) the dirty bitmap from the previous run is - allegedly - invalid and is cleared. And so the backup (currently shy of 600GiB) takes forever.

Since there are no errors, I have no idea where to start debugging this.

How can it be that the dirty bitmap is invalid every single time, after less than two hours?

Any ideas?

Thanks!

What type of backup are you running? As soon as the VM is stopped the dirty-bitmap will become invalid. A "stop mode" backup for example won't be able to use any dirty-bitmapping because it will stop the VM before each backup.

Snapshot and the VM has been up for several days without interruption.

ah, yes. There are two backup jobs for this VM to different targets (I am in the process of migrating from my old PBS installation to a new one).

I have disabled one and will report back.

The normal backup works.

Thanks for your help!

Then it's expected, because the bitmap is only valid for a given target (and we don't save multiple bitmaps).

Hi, and is it possible to do so? Save one bitmap by (pbs,namespace) couple?

More and more PVE users have multiple target PBS...

Regards,

Christophe.

IMHO it's better to use PBS sync to copy backups from one PBS to another instead of doing backup of the VM multiple times, even if PVE could use multiple dirty-maps for a few PBS. Less load on source PVE, less load on VMs and you offload the responsibility of doing the second backup replica to the backup server.

Well, it's also my point but some users asked for such a feature.

Even though this is an old thread, I think this is a great idea to support dirty-bitmaps per target: This would allow 2 (or more) completely separated and fully independent backups sets.