Configure StarLvmPlugin cluster lock timeout

[rhp]

We are in the process of moving our workflows from vSphere to Proxmox.
Part of our workflows rely on creating snapshots and reverting to a "clean state" after running a suite of tests by rolling back to said snapshot.

We are using StarLvm (1.1.41) on Proxmox VE 9.2.4 to store the VMs and snapshots

On Proxmox we regularly see that creating, deleting, rolling back a snapshot fails on a timeout trying to acquire a cfs lock for this underlying storage:

Code: Select all

trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
trying to acquire cfs lock 'storage-CITA' ...
TASK ERROR: cfs-lock 'storage-CITA' error: got lock request timeout

It seems that this timeout is set to 10 seconds. However, we observe that typical snapshot operations take 5 to 15 seconds to complete, which means that this timeout is relatively short in comparison.

Is there a means to increase the cluster-level lock on the storage to set it to a higher value?

Kind regards,

Ronald

[rhp]

A related question: when this timeout happens and the task gets aborted, this leaves the VM in a locked state.
The only way to remove the lock is using root-level credentials, which we do not want to assign to just everyone.

Would it be possible to leave the VM unlocked if the cluster-lock cannot be obtained?

[yaroslav (staff)]

Hi Ronald,

I've forwarded your request to the dev team. Hope to see improvements in future versions of the plugin.

[yaroslav (staff)]

Hi,

Could you please collect the logs /opt/starwind/starwind-proxmox-plugin/collect_pve_logs.sh (please indicate the timestamp when the incident happened) and upload them here ? Thanks!

[rhp]

Hi Yaroslav,

These logs would be massive, and honestly, I do not think this provides any information that is not already clear.
Because of the growing LVM archive folder, LVM operations are slow sometimes, and we simply need a longer timeout value.

We have this running at our side with the changes I sent in a PM, with a timeout of 60 seconds, which looks like enough to fix this issue (for us).