How to climb vm, after failure of 2 nodes out of 3, running on GlusterFS ?

Alexandre Aguirre

Well-Known Member
Apr 4, 2018
30
4
48
39
Passo Fundo - RS, Brasil
Hello guys !

I'm doing some labs with storage glusterfs, in the tests I have 3 nodes, everything working perfectly, but I had a problem simulating the failure of 2 nodes, the vm doesn't start because it's not able to recognize the datastore......... .I'm still new to GlusterFS, if anyone can help me I'll be grateful!

Follows images:erro_gluster.png
erro3_gluster.png
 
Hi,
your Proxmox VE cluster is not quorate (see here) and thus read-only if two out of of three nodes are down.
 
Hi,
your Proxmox VE cluster is not quorate (see here) and thus read-only if two out of of three nodes are down.

Hello good day !

Thanks for the feedback, could you just help me out? I read the documentation that you gave me, but I didn't understand it very well, because even though I was running pvecm expected 1, it still didn't work.............. Would it refer to QDevice?
 
You should not run pvecm expected 1, because the two offline nodes won't see the changes to the cluster state - making it inconsistent! If you are running a 3-node cluster, at least 2 nodes need to be up for the cluster to work. If that is not possible, you shouldn't use a cluster. A QDevice only makes sense for an even-sized cluster.
 
You should not run pvecm expected 1, because the two offline nodes won't see the changes to the cluster state - making it inconsistent! If you are running a 3-node cluster, at least 2 nodes need to be up for the cluster to work. If that is not possible, you shouldn't use a cluster. A QDevice only makes sense for an even-sized cluster.
Thanks for the quick turnaround!

So in this case, if I lose 2 nodes, I won't be able to somehow temporarily raise my vm?
 
Thanks for the quick turnaround!

So in this case, if I lose 2 nodes, I won't be able to somehow temporarily raise my vm?
Yes, because the cluster won't be quorate from PVE/Corosync's perspective.