2 Nodes Cluster - Help :(

bartounet

Member
Dec 11, 2022
31
1
13
Hello, As part of a home lab,

I currently have a Proxmox server with VMs.

I purchased another server to give me more flexibility and security.

I've been researching the best solution for using these two servers for quite some time.

I initially wanted to start with a hyperconverged Proxmox + CEPH cluster, but I quickly came across the concepts of Proxmox quorum and Ceph qorum.

In the end, I'm a little lost.

What is the best solution for creating a cluster (at least with administration on the same interface)?

- 2 Proxmox servers + CEPH? ( + Corosync on RPI?)

- 2 Proxmox servers + ZFS pool and replication? (Does qorum need?)



Can you explain to me again the concepts of quorum for Proxmox and also for CEPH (they're not the same??).

Ultimately, if I don't want to use HA, is a quorum of 3 required?
 
For the proxmox cluster, you can run 2 servers with the quorum on a raspi, lxc or VM .... ZFS replication worked for me very well, the downside is, that you can have a data loss of 1 minute (depending on your sync interval). The good part is, its dead simply and you can't kill it ... i run that setup for 2 years or so without any problems (with 10G and 1G network connections), would recommand it for personal use case.

For ceph, you should use 3 servers and a good network connection between the servers and when something went wrong, you should know what you do and have a good backup strategy, since ceph is not so easy to handle (search the forum, many people have issues and problems to fix it).

The quorum for proxmox is to avoid a split brain situation, when two server loose the connection to each other, who should be the working one? With three votes, the two wins, who see each others.

For ceph its more a how much copy of file do you want to be saved an the hosts.

When you only want the cluster for the single webgui, you could try the DataCenter Manager.
 
Thanks for answer
it is important to have qorum about 3 node (with raspi) when HA is not active ?

Other questions
Regarding ZFS and HAThe Quorum of 3 is important ?
yes, I imagine, because it's HA.

If we replicate the VM from pve1 to pve2, for example,
If we lose PVE1, does the VM automatically restart on PVE2 from the replica?
But what happens when PVE1 restarts?
Does replication take place in the opposite direction?
 
Last edited:
no i dont understand sorry.
Actually i am testing 2 pve nodes on cluster whith local disks
if i stop pve1, vm on pve2 stays to work fine, not on read only
 
Last edited:
What you think about this approach
give 2 votes at nodes
(( pve1 has two votes; pve2 has one. ))

What do you think happens when his "pve1" dies? "pve2" is not usable in that case. (Not without some manual intervention for troubleshooting.) That's just not what I would call a working cluster!

But yeah, under some circumstances you can fake several system details, if you know what you are doing and what to expect. Beginners do NOT have this experience usually...
 
Last edited:
If we replicate the VM from pve1 to pve2, for example,
If we lose PVE1, does the VM automatically restart on PVE2 from the replica?
But what happens when PVE1 restarts?
Does replication take place in the opposite direction?
when you have HA configured for the VM, it will automatically start on pve2 after some timeout. When pve1 restarts, the VM will be migrated back to pve1. yes, replication automatically switch direction. So you only have to configure it once. With replication you could also live migrate VMs between the servers.

When you don't want to use HA, checkout the DataCenter Manager, where you could also migrate VMs between different servers.
 
  • Like
Reactions: Johannes S and UdoB