MTU on vm's suddenly changed when started on proxmox 9 host.

R

Ronny Aasen

Active Member
Proxmox Subscriber
Mar 15, 2018
18
7
43
Norway
Hello.

I have solved my issue, by setting mtu=1500 on all nic's in all vm *.conf files. using sed inline. This is just to understand if there is something strange in my enviorment. or if this is as expected.

if a vm, started on a proxmox 8.x node, is migrated toa proxmox 9 host. and then stopped and started again, (os reboot is not sufficient). the network interface in the vm have mtu 9000 after boot.
eg:
2: ens18: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc pfifo_fast state UP group default qlen 1000
link/ether c2:1f:3e:91:7b:52 brd ff:ff:ff:ff:ff:ff

settingt the mtu explicitly on each vm's network cards, and restarting the vm fixes the problem. one can also configure it inside the os's network settings.

I assume this comes from the change in default behavior for virtio nic's in proxmox 9.0 from 1500 to whatever the bridge have. https://bugzilla.proxmox.com/show_bug.cgi?id=6802
my vlan aware bridge vmbr1 on a lacp bond have mtu 9000, since i have some vm's that participate in a storage network.

Having many vm's cease to function on a restart came as a suprise. perhaps a notification in the pve8to9 script that mentioned this if you have mtu larger then 9000 on any bridges would be a nice head's up. and/or in the documentation about the upgrade.


thanks for proxmox!
Kind regards
Ronny Aasen
 
Ronny Aasen said:
Having many vm's cease to function on a restart came as a suprise. perhaps a notification in the pve8to9 script that mentioned this if you have mtu larger then 9000 on any bridges would be a nice head's up
Click to expand...

This should already be the case - is there any possibility of re-checking whether the script did not output a warning in your case?
 
these clusters are not possible to recheck, they are all fully migrated.
I will keep it in mind for future migrations, to check that.

From memory, for the clusters i updated myself, there was no warnings left in pve8to9 when i did my upgrades.

But I do see the output
"INFO: Checking for VirtIO devices that would change their MTU..."

I had drained the host for all vm's before starting on the migration (and running the pve8to9 tool) tho. will the mtu check consider other vm's in the cluster, that will eventually be migrated back to this host. Or does it only consider vm's actually on the host ?

Best regards
Ronny Aasen
 
Ronny Aasen said:
I had drained the host for all vm's before starting on the migration (and running the pve8to9 tool) tho. will the mtu check consider other vm's in the cluster, that will eventually be migrated back to this host. Or does it only consider vm's actually on the host ?
Click to expand...

It only considers VMs on the current host, so that might be the reason why you didn't get a notice from the migration tool.
 
You must log in or register to reply here.
Top Bottom
Back