Node replacement after hardware failure on a cluster

[agh]

Hello,

I've a production cluster composed of 4 nodes :

• node1
• node2
• node3
• node4

I had an hardware failure on node4. I've to change the node, and to reinstall Proxmox on this node.
Of course, I migrated all the VMs which were on node4 to other nodes.

As I am on a production cluster, I want to be 100% sure of the procedure to follow. Could you tell me if I'm wrong ?

1. apt-get update && apt-get upgrade on all nodes 1, 2 and 3
2. fresh install of Proxmox on node4, with same IP and hostname
3. apt-get update && apt-get upgrade on node4, to be at the same level as other nodes
4. on node1 for instance (or 2, or 3), do :
   pvecm add node4 --force
5. reboot node4

Is it correct ?
If not, what should be the procedure ?

Thanks a lot.

 

[tom]

no, the command for adding the node is wrong.

you need to:

> pvecm add IP-OF-ONE-OF-YOUR-EXISTING-NODES --force

see http://pve.proxmox.com/wiki/Proxmox_VE_2.0_Cluster

and a reboot is also not necessary, I assume you already rebooted after updated your box and kernel.

 

[agh]

Hello,

Thanks for your answer.
I'm in trouble. Murphy's law you know...

I've another node which failed. So, I've lost the quorum (only 2 / 4 nodes up).
And, without quorum, i can't succeed in reinstalling new nodes and adding them into the cluster.

Is there a trick ?
It's really important as it is a prod cluster.

Thanks a lot

 

[jleg]

I've another node which failed. So, I've lost the quorum (only 2 / 4 nodes up).
And, without quorum, i can't succeed in reinstalling new nodes and adding them into the cluster.

Is there a trick ?

tell cluster to use a quorum of 2:
pvecm expected 2

 

[agh]

Thanks a lot !
It worked.