I have updates on my installation stability. With the upgrade to pve-qemu-kvm 11.0 from 10.2.1, proxmox 8 kernel (6.8.12) was having trouble with i/o threading.
Suddenly every vm was queueing I/o, to the point that global iodelay was greater...
What did you switch to?
I'm still on 8.4 with an updated kernel ( 6.8.12-22-pve ). Lots of nfs connections on the proxmox host to TN (a vm on this host with the onboard sata controllers passed through). No issue as far as stability. I've been...
It's good that those parameters worked, but disabling ncq is reducing performance. I think the work around is masking the issue rather than solving the problem entirely.