How to disable quorum mechanism in a cluster ?

[shodan]

Hi,

I have setup 4x of my proxmox servers as a cluster so that I may hopefully move VMs around with ease.

However, when I have only one server powered, the quorum mechanism blocks all operations.

I understand that commercial IT does not have a concept "a server turned off on purpose" so this is an entirely new in computer science.

So I would like to explore the possibility of, a proxmox cluster with no quorum mechanism ?

Or what is the proper way, if any, to do that ?

I did find the command

pvecm e 1

Will remove problem with quorum, but it seems this only takes effect until the next reboot ?

Is there a proper way to make that permanent ?

or should I just put that command in

/etc/rc.local

?

 

[leesteken]

A PVE cluster is based on quorum; nothing to do about that except force the expected number of votes (and possibly get in all kind of trouble): https://forum.proxmox.com/threads/c...art-without-quorum-and-expected-votes.143836/

This is the upcoming way to manage multiple clusters (or multiple nodes without a cluster): https://forum.proxmox.com/threads/proxmox-datacenter-manager-first-alpha-release.159324/

 

[aabraham]

Hi,

I have setup 4x of my proxmox servers as a cluster so that I may hopefully move VMs around with ease.

However, when I have only one server powered, the quorum mechanism blocks all operations.

I understand that commercial IT does not have a concept "a server turned off on purpose" so this is an entirely new in computer science.

So I would like to explore the possibility of, a proxmox cluster with no quorum mechanism ?

Or what is the proper way, if any, to do that ?

I did find the command

pvecm e 1

Will remove problem with quorum, but it seems this only takes effect until the next reboot ?

Is there a proper way to make that permanent ?

or should I just put that command in

/etc/rc.local

?

Why are you trying to turn off the Quorum mechanism?

 

[shodan]

3 out of 4 servers are offline
I see that I could bypass this problem by hosting a "quorum" device in an LXC container of each of the servers
But it seems easier to have each server permanently have a vote expectation of 1

The only reason why these servers are clustered together was for ease of send one VM or CT from one server to another with minimal downtime and mostly, ease of a single click. That's not something I have actually tried to do yet.

I am considering instead figuring out a script that can turn any VM or CT into a file, then copying that file to another proxmox server and re-starting the VM/CT there. Or possibly moving all disks to iSCSI disks on the network.

 

[leesteken]

I see that I could bypass this problem by hosting a "quorum" device in an LXC container of each of the servers

That's not how a QDevice works. This might work with a two-node cluster but not four.

But it seems easier to have each server permanently have a vote expectation of 1

Yes, but also easy to shoot yourself in the foot and end up with a split-brain cluster.

I am considering instead figuring out a script that can turn any VM or CT into a file, then copying that file to another proxmox server and re-starting the VM/CT there.

That's simple: backup the VM or CT and scp it over the network. Or run a PBS in a container on the "main" node and restore to other nodes when you want to.
Or try the new multi-cluster/node manager...

 

[waltar]

Maybe would make more sense and be better to run your 4 pve nodes as 4 single nodes without clustering to reach your goal.
Then you can manually copy over images or still have shared storage like nfs to make that even easier too.

 

[mfed]

In theory, you should be able to modify corosync configuration (/etc/pve/corosync.conf) and assign to the node that you want to keep enough of votes so that the single node will be the quorum with no regards to the other ones (e.g. set quorum_votes to 4, the other 3 nodes with quorum_votes of 1 in total will have 3 votes).
If you are going to mess with it, please make sure to read the doc how to edit the corosync configuration properly at https://pve.proxmox.com/wiki/Cluster_Manager

 

[waltar]

The other theory behind that is when your main node with high quorum dies and you start up the others to run your services (from last backup) that won't work as /etc/pve/ is readonly and you cannot change quorum votes back ... so it's a little dangerous ...

 

[mfed]

Yes, that's when you would need the 'pvecm expected 1' for the recovery. Better IMHO than using 'pvecm expected 1' permanently...