[SOLVED] Two node cluster with a third quorum

P

PatricF

New Member
Sep 3, 2014
4
0
1
Hi,

I'm just trying out Proxmox for the first time and had some questions.
I'm going to start with just one node and in a couple of weeks/month add a second node. I will make a cluster so I can manage these two with the same webUI.
However I don't like the fact that I won't be able to take down one node for maintenance that easy without getting a split-brain scenario.

I have a separate server on the same network running VMware with a bunch of Linux VMs. I thought that I could use one of them to act as a third quorum so one node can go down without getting a split-brain scenario.
Is this a good idea or will there be other complications when the second node is down even if the first server has quorum?
In the near future I will add a third server and so on, so I will eventually not need the third quorum-only node.

What are your thoughts and how would you configure this scenario?

Thanks,
-Patric
 
Last edited:
m.ardito said:
Hi!

see http://pve.proxmox.com/wiki/Two-Node_High_Availability_Cluster
and in particular: http://pve.proxmox.com/wiki/Two-Node_High_Availability_Cluster#Quorum_Disk_Configuration

Marco
Click to expand...

Thanks for the quick reply, I will read trough the link. At the moment I'm not really interested in HA, I just want to manage all my nodes with the same webUI.
Maybe there's no problem if that's all I want to do to start with?
Of course I will configure HA eventually but to start with I'm only going to use the cluster for central management.
Is HA needed to configure a cluster (only interested in central management)?

Thanks,
-Patric
 
you need quorum to manage the cluster (without HA),
becuase they are no master in proxmox, so any node can write vm configuration.

In 2 nodes cluster, if 1 node is down, you'll be enable to start vm, edit config,...


I think that it could be possible for your setup, to create a proxmox vm on your vmware server, just to have the quorum.

Alternative could be to use a shared storage iscsi lun, to act as third node quorum.
 
well with no HA, restarting a vm which was on a failed node is a manual action, so, if you have a split-brain situation in this setup... the brain is yours! :-D all depends on it!

You could restore a vm backup on another node, better with a new VMID, but you have to be careful: the original node could still have that vm, and the vm could be set to auto-start, so when the repaired node comes back in the network, and thus in the cluster, you will have the same vm, under two VMID (the original one on the repaired node, the new one on the node where backup was restored), same MAC addresses, same hostname, ssh keys,...everything!
you have to manually take care of not restarting the same vm twice, then.

HA (properly working) takes care of this. But when you have a cluster, HA is only an option.

But even without HA, a cluster always has quorum. And a cluster (even no HA) without quorum is locked.

A 2 node cluster has an expected quorum of 2, and will loose quorum when 1 node fails (1 left vote is no quorum)

So you can
- set your 2 node cluster to "expect" only 1 vote, and so your cluster will never fail, never be locked.
- Or (better) you can use the third "voter" (a node or a quorum disk), and use an expected quorum of 2 (stndard for 3 nodes), so when 1 node fails, cluster sill has 2 votes over 2 expected, and the clsuter keeps fully operational..

Marco
 
Thanks a lot for all your replies, it's much appreciated. I think I got the information that I needed to continue with this project.

-Patric
 
You must log in or register to reply here.
Top Bottom