Cluster with GlusterFS doesn't work fine

Romain DARBON

New Member
Jan 23, 2019
3
0
1
27
Hi,

I would like to configure a 2 nodes cluster with GlusterFS stockage. I know that it isn't a good Idea and i sould have 2 nodes but I doen't want HA, I only would like to can manually start a VM on the second node if the first fail.

So I configured my cluster with 2 nodes, and I created a glusterfs storage between both nodes. I can show it in both nodes and I can configure a VM with this storage.

But I meet a problem when one of nodes fails. The storage switch to "Active : No" in the promox interface whereas it's active in the shell (I check with the command "gluster volume status HDD-GLUSTER".

First i thinked that it's because i hadn't the quorum so i set the quorum votes to 2 in the first node. So when I stop the second node, thet first have still the quorum. But the gluster volume isn't active in proxmox.

Thanks a lot for your help,
Sorry for m'y bad english i'm french,
Best regards,
Romain.
 
I would like to configure a 2 nodes cluster with GlusterFS stockage. I know that it isn't a good Idea and i sould have 2 nodes but I doen't want HA, I only would like to can manually start a VM on the second node if the first fail.

So I configured my cluster with 2 nodes, and I created a glusterfs storage between both nodes. I can show it in both nodes and I can configure a VM with this storage.

But I meet a problem when one of nodes fails. The storage switch to "Active : No" in the promox interface whereas it's active in the shell (I check with the command "gluster volume status HDD-GLUSTER".

First i thinked that it's because i hadn't the quorum so i set the quorum votes to 2 in the first node. So when I stop the second node, thet first have still the quorum. But the gluster volume isn't active in proxmox.

Such a setup is not recommended at all; if you want to continue working when one node fails install at least a 3-node Cluster. Put a third node to your cluster, can be a low-end machine (you have it just for quorum then)

However, if you really want to try it with a 2-node cluster:

You have to set Quorum to 1 in such a case:
Code:
pvecm e 1

And observe if quorum is (and remains) ok by
Code:
pvecm status
 
Hi Richard,

I think it's more reasonable to work with a third node. I tried with "pvecm e 1" but nothing change in "pvecm status".

When I shutdown one node, the last one change state to "Quorate : No".

It's very strange !

Thanks for your help,
Romain.
 
Hi Richard,

I think it's more reasonable to work with a third node. I tried with "pvecm e 1" but nothing change in "pvecm status".

When I shutdown one node, the last one change state to "Quorate : No".

Code:
pvecm e 1

has only temporarily effect; i.e. as soon as something changes in cluster state it's set back to default (which is 2 in case of a 2 or 3 node cluster).