Working SDN completely broke after applying a change

[blackline]

Hi, we have 2 PVE clusters in prod

1. mother cluster with 3 nodes
2. worker cluster with 3 nodes

They are more or less identical, have similar SDN configurations etc.

Today we made an addition to the SDN configs on both clusters. The mother cluster did fine, but for the worker cluster all vlan based communication (the ones defined in SDN) stopped working! The only thing that helped was

1. create a new vlan with on each node
2. create a new vmbr pointing to the new vlan
3. reconfigure 200 VMs to the new vmbr

We are going to not use SDN moving forward, this is a serious bug and we can't trust it in this state.

If there are things we could do to help debug this, I am all ears.
Is anyone else using SDN in production?

 

[shanreich]

There was bug related to bonds on non-vlan aware bridge with the newest version, which should already have been fixed in the meanwhile with [1] and the respective package should be available on testing for now.

Is it possible that this was the cause in your case?

[1] https://lore.proxmox.com/pve-devel/9bad9706-d9c9-433f-85a4-d057a6b67e5e@proxmox.com/T/#t