Cluster with permantly expected votes to 1?

[Edox]

Hi,

I use two desktop PC's at home, both running Proxmox.

I put them in a cluster, because then it's easy for me to upgrade proxmox: remove all VM's to the 2nd machine, upgrade the first one, then put all VM's back. And nobody in the household notices anything interrupting. Works very well.

However, because of energy consumption, I would like to turn off the 2nd Proxmox device when I am not using it (99% of the time). However, when I do so, the qurum is not enough, and I can't even login to the still power-on device. Also, a few days ago there was a power interruption, and the single Proxmox didn't want to start all the VM's because the qurum was too low.

Is there a way to have a cluster, and permanently keep expected votes to 1?
Or should I have a 3rd (low power) device allways on to keep the quorum high enough? Any suggestion what this could be? RPi?

I know this is not the recommended setup, but it's for home, and I don't use any shared storage.

Thanks.

 

[Dunuin]

By design a cluster needs at least 3 voters and the majority (so 2 out of 3) have to be running all the time. It's not recommended to run your two nodes with expected votes 1.

And yes, a raspberry pi could run as a qdevice and you can even use it as a third PVE node: https://github.com/pimox/pimox7

 

[untitl3d]

@Dunuin (or anyone) could you please elaborate what exactly could go wrong and why it's not recommended?
I have a pfSense running on Proxmox and the whole network is down until it boots up so I have the same issue with expected votes.

 

[LnxBil]

@Dunuin (or anyone) could you please elaborate what exactly could go wrong and why it's not recommended?
I have a pfSense running on Proxmox and the whole network is down until it boots up so I have the same issue with expected votes.

A split brain can happen including all possible side effects like data corruption.

Depending on your HA needs, I recommend building two clusters: one for the HA firewall and one for the HA rest. You should also have a highly available DHCP and DNS server in each cluster so that you will have networking if one of the clusters fails.