Unstable cluster (multicast issue?)

[Rizki Rivai]

We have 8 node proxmox 5.2.1 cluster. We experience cluster disconnection when we clone vm. We also have fail to connect to storage.
Many thread suggest to check at multicast isu. We followed https://pve.proxmox.com/wiki/Multicast_notes#Linux:_Enabling_Multicast_querier_on_bridges and we have some node shown > 5% loss. Some time other node show > 5% loss.
We also notice that some node have Can't get key totem.interface.0.mcastaddr. Error CS_ERR_NOT_EXIST

we disable igmp snooping on switch but omping test still shown roughly same number.

any suggestion what to check?

our switch is mellanox using cumulus

 

[Rizki Rivai]

restarting corosync resolve CS_ERR_NOT_EXIST. But omping result still not stable. Sometime it shows loss 1% sometime can be as high as 19%.
I am still waiting on disconnected issue as it happen quite randomly.