All VMs stuck after restore other VM over FC storage

[fips]

Hey Guys,

I have a really stupid issue, I hope you can help me.
My setup: I have 3 proxmox hosts (all 4.1) they use a shared storage connected over FC, backup is made to another NFS storage.
All VMs running like a charm except I have to restore a VM.
In that case the other VM's get errors like: "task jdb2/dm-0-8:212 blocked for more than 120 seconds", "Buffer I/O Error on device dm-0", "end_request: I/O error dev vda" and the filesystem of the VM's will be remounted RO...
So I have to reboot all VM's and run a "fsck -y" otherwise they would not start.

Do you have an idea how can I prevent that VM's block after a restore?
Attached you can find the error messages.

THANKS

 

[fireon]

Hmm, I/O Error. Can you to an filesystemcheck please? On the VM and on the host? On the Host "touch /forcefsck". On the VM with an livedisc.

 

[fips]

Well I have to run a filesystemcheck on the VMs otherwise they can't start.
But I need a solution for prevent the VMs of blocking during/after a restore.

 

[LnxBil]

This error seems like your running VMs do not get enough I/O power/time when the backup runs. This can be scheduling problem.

What SAN do you use?
Do you use clustered LVM ?

 

[fips]

I use Open-E DSS6, and no I don't use a clustered LVM.

 

[LnxBil]

I use Open-E DSS6, and no I don't use a clustered LVM.

iSCSI or NFS? What about saturation of ethernet? You can monitor the traffic on your switches?

 

[fips]

Neither nor...
all 3 hosts are connected via fibre channel.

 

[LnxBil]

Then what are you using for shared storage? As far as I know there is only clustered LVM available as shared storage on FC.