Storage replication and VSS

[mdenadal]

Hello,
I'm using zfs storage replication across nodes with great success.

Is it possible to disable "qm guest fsfreeze-freeze" during replication?

I have performance issues replicating Windows VMs using VSS enabled services (especially SQL Server).
I know this can introduce consistency issues but the performance drop caused by VSS quiesce in some cases are unacceptable.

Thanks
Massimo

 

[dcsapak]

this is currently not configureable, but you could not use the qemu-guest-agent

i am not sure, but maybe there is a way to tell windows not to do vss when the guest agent requests it?

 

[mdenadal]

Hi Dominik,
thanks for you answer.
Sure, I can disable the agent, but this is not a solution.
I need to take "quiesced" (fsfreeze-freezed) snapshot during low pressure work hours.
Disabling the agent I can't take quiesced snapshot anymore.

Do you think I can ask a feature to disable fsfreeze during replication for a next release?

 

[dcsapak]

you can open a feature request, but i can make no promises if and when it will be done https://bugzilla.proxmox.com/

 

[mdenadal]

Ok, I'll open a feature request.
Thank you.

 

[jsterr]

Is this feature available in 2023?

 

[AlexHK]

I may have run into a related issue, but I am not too sure since I am not very familiar with VSS.

I am replicating a Windows VM to 3 other hosts every 5 minutes. It usually takes only a couple of seconds (~7 sec) for each host/replication run.

Within the VM I run a backup software called Macrium Reflect creating image backups of the three drives at around noon time daily.

Once in a while I get errors, today it backed up 1 out of 3 windows drives, failed on the other 2 afterwards. I guess this is happening when by coincidence the replication happens at the exact moment Macrium Reflect tries to take a VSS snapshot of a drive.

https://knowledgebase.macrium.com/display/KNOWX/VSS+error:+VSS_E_SNAPSHOT_SET_IN_PROGRESS

VSS_E_SNAPSHOT_SET_IN_PROGRESS indicates that another application has an open volume snapshot on the volume being imaged. It is only permitted to have one open volume snapshot on the same volume at any one time. This could be caused by another backup product or by Windows updates. If backup runs at the same time as Windows updates then this could be the cause of the problem.

The feature request is still open after 6 or 7 years, so I guess for now the obvious solution (for me) is not to run inside-VM backup and replication a the same time.