Permanently Disable Cluster Quorum Requirement

[JBdrakaris]

Good afternoon,

As the title states, is there a way to permanently disable the cluster quorum requirement?

We do not use shared storage or HA, so my understanding is that the quorum requirement is not necessary.

Thanks!

 

[Duckdave]

Good afternoon,

As the title states, is there a way to permanently disable the cluster quorum requirement?

We do not use shared storage or HA, so my understanding is that the quorum requirement is not necessary.

Thanks!

have a look at
pvecm expected <number>

i use this to get a 2 node cluster without ha and local zfs storage.

 

[danielb]

You still need to be quorate to prevent split brains on /etc/pve filesystem

 

[JBdrakaris]

Good morning,

We are testing PVE as a possible path forward for our current infrastructure, with nodes in different geographic sites. The appeal of PVE for us is the ability to view all the systems from one panel and perform offline migrations.

As I understand it, "pvecm expected" is temporary and resets once all is good.

I also believe I understand the perils of a split brain on the /etc/pve filesystem. That being said, since we are not running shared storage or HA, it seems the risk of differences on the filesystem is minimal between the various nodes.

We had an instance that a site lost connectivity for a few hours. We took the opportunity to bring the node down and perform offline maintenance. Once connectivity was regained, the node had issues reconnecting to the cluster, and refused to automatically bring the vms back online. At that point we rebooted the entire cluster, which resolved the issue. These being test machines, this was an option.

If these had been our production servers, rebooting the entire cluster for one node would not have been an option. We do need the nodes to immediately start the vms on them regardless of quorum.

Which leads back to the original question, is there a way to permanently disable the cluster quorum requirement?

Or am I not understanding something?

As always, I'm open to suggestions.

Thanks!

 

[danielb]

I don't think there's a supported way to do this. For this use case, it's recommended to run independant (non clustered) Proxmox. There's just too much risk of split brain and corruptions if you operate on a non quorate node. This is not specific to Proxmox though.

 

[JBdrakaris]

OK. That's what I was thinking.

Could the problem be the slower internet connections between sites?

To partially overcome, I was thinking of creating a cluster for each site, and then mounting the same nfs backup folder on each site. That would allow easy manual migrations... but not the unified overview of all the systems (which is what we really wanted).

For our smaller sites with 3 or less nodes, I believe quorate automatically adjusts.

All that being said, is there a way to get the vm's to just boot without waiting on quorum? They are independent, not ha, shared, etc, so there is no need for them wait on quorum. I was hoping for an option in the config file such as: " DoNotWaitOnQuorumToBoot: YES "

At the end of the day we just need them to spin up without intervention, such as would happen if they were not in a cluster.

 

[JBdrakaris]

I think I found it:

https://www.systutorials.com/docs/linux/man/5-votequorum/

two_node: 1

For anyone wanting to do this, make sure you understand all the implications...

 

[JBdrakaris]

Nope, that ain't the answer...

NOTES: enabling two_node: 1 automatically enables wait_for_all. It is still possible to override wait_for_all by explicitly setting it to 0. If more than 2 nodes join the cluster, the two_node option is automatically disabled.