Cluster HA - 2 proxmox PVE 4.3 nodes and quorum disk

[djahida]

Hello,
In Proxmox 3.x we were able to create an HA cluster with two Proxmox nodes and quorum disk.

Is it possible to have this architecture in Proxmox 4.x?

Thanks

 

[udo]

Hi,
no!

Udo

 

[djahida]

Thanks Udo for the reply
can you tell me why?

 

[tom]

there are some posts and threads regarding this questions, please use the search function of the forum.

 

[djahida]

ok. thanks

 

[djahida]

So, is it possible to install cluster three node as
- two nodes have the same (RAM, size disk, CPU, ....)
- but the third node have less performance on RAM, CPU but the same size disk with the two others

thanks

 

[morph027]

Sure. If you're brave enough, your third one can be a VM on some other host to. You do not even need to deploy vms on this host, so disk isn't critical at all. It's just for clean qorum state (this is no Proxmox "issue", almost all clustered solutions need un-equal numbers of members to avoid split-brain).

 

[djahida]

thanks for the reply
but if I have to use DRBD in the cluster I must have the same size storage on the three nodes. I don't think that it is possible with a VM

 

[morph027]

You can use DRBD on 2 nodes while having the cluster on 3 nodes.

 

[djahida]

in these case if I have a failure on the cluster is there any risk to have all VM on one node migrating to the third node (VM) ?
I mean, how configure these cluster ( 3 nodes ) to use only the two phisical machine in case of failure

 

[morph027]

You can specify HA groups https://pve.proxmox.com/wiki/High_Availability

The HA group configuration file /etc/pve/ha/groups.cfg is used to define groups of cluster nodes. A resource can be restricted to run only on the members of such group.

 

[djahida]

thank you so much
I used the pve version 3 and the notion of groupe HA was not existe
juste an other questoin please:
when I simulate a faliure on one node like 'systemctl poweroff -f', the fencing( I use for the moment the default way: watchdog) reboot the node
where configure this option to maybe stop the failure node not reboot it