Proxmox 2 nodes cluster - unable start VM after one node fails

[derata]

Hello, I have cluster with two nodes. After transfer HW servers to the new room one node stopped working (second in cluster). Now is impossible to start this node. Bad motherboard.
When I start first node. No VM start. Proxmox say "waiting for quorum ..." when turning on VM.
When i tryed check cluster pvecm status live node sayed "Cannot initialize CMAP service" and all other command pvecm say the same.
In GUI a still see second node. I donť understand, why I cannot turn on VM on live node.
Thanks for help.

 

[dietmar]

If you are sure the other node is offline, you can maually set expected votes to 1:

# pvecm expected 1

You should get quorum after that.

 

[derata]

If you are sure the other node is offline, you can maually set expected votes to 1:

# pvecm expected 1

You should get quorum after that.

I tried it before. But with same error "Cannot initialize CMAP service"

 

[dietmar]

Is corosync running?

# systemctl status corosync.service

 

[derata]

Yes it running.


I used and I was able delete cluster information on first node and start VM.


rm /etc/pve/corosync.conf
rm /etc/corosync/*
killall pmxcfs
systemctl start pve-cluster

But now i simulate this situation again, because I will want use 2 node cluster in my production enviroment. I installed proxmox to 2 testing HW. On the first node I created 2 VMs (local and NAS storage) a then a created cluster with second node. Then I turn off second node and when I stop VMs on first node and trying start VMs again - it ends with same error "waiting for quorum". Create backup is impossible too.

Is it normal situation that when one node is shutted down, than is imposible start VM on other node? I don't understand this. If one node in my prduction enviroment failed, then I have problem on everywhere. Than is better solution for me - don't use cluster for me.

 

[dietmar]

Is it normal situation that when one node is shutted down, than is imposible start VM on other node? I don't understand this. If one node in my prduction enviroment failed, then I have problem on everywhere. Than is better solution for me - don't use cluster for me.

The cluster is a quorum based system. On a two node system, you lose quorum if the other node is down. This prevents split brain situations.

But normally you can simply set expected votes and you are done.

 

[derata]

OK. Very thanks. And last question - if I set up third node (some older hw) just for quorum. I am OK if one main nodes failed?