Hardware for Proxmox High Availability Cluster

[skrod]

What is the recommended hardware for a proxmox high availability cluster?

From my understanding. You have two options. 2-node HA cluster and a 3-node (recommended) HA cluster.

It's not explained well as to what each node's role is.

"If you run HA, only high end server hardware with no single point of
failure should be used. This includes redundant disks (Hardware Raid),
redundant power supply, UPS systems, network bonding"

Is it required to have the exact same server? Dual Xeon, 36GB ram, 4x480GB SSD HW Raid10? times 2-node or times 3-node?

On a 2-node cluster can you use a Dual Xeon, 36GB ram, 4x480GB SSD HW Raid10 and the same server for server2 with 4x1TB SATA HW Raid10? Obviously losing performance if server1 goes down and server2 takes over? But regaining performance when server1 returns.

On a 3-node cluster can you use a Dual Xeon, 36GB ram, 4x480GB SSD HW Raid10 and the same server with 4x1TB SATA HW Raid10? For server 3, same exact hardware (processor/ram wise) but 2x1TB SATA HW Raid1 or 2x480GB SSD HW Raid1, can you use a smaller drive? (2x120 SSD HW Raid1) Can you use less ram? for server3?

From my small amount of research, it's not quite clear what each servers roles are and what each server hardware requirements are. Could anyone please explain?


Thanks in advance!

 

[wolfgang]

Hi,
first of all 2 nodes are possible, but no really recommended if you need Ha.
The server must not be the same HW.

You must create Failover Domains,
to define where the Vm's should be migrated in case of node failure.
You should have in mind the it must possible to run the vm on the other nodes if a node crash.

 

[skrod]

Wolfgang,

Thanks for the response. What exactly do you mean "the server must not be the same HW."?

This is the reason why I wanted to know each nodes role in a 3-node HA cluster. So I know where high end hardware is needed.

 

[mir]

As in 'does not need to'. However, it would probably be a good idea to have at least not a mix of Intel and AMD. The next thing to be concerned about is that CPU's are of same version, eg. Xeon v2 on all nodes. Reason: To ensure online migration all VM's must follow emulation of lowest common denominator.

 

[udo]

Wolfgang,

Thanks for the response. What exactly do you mean "the server must not be the same HW."?

This is the reason why I wanted to know each nodes role in a 3-node HA cluster. So I know where high end hardware is needed.

Hi,
it's depends from your storage. The nodes must be able to access the storage where the VM-disks are stored.

But perhaps the third node must not, because it's only an small server for quorum (I have on two cluster an hp microserver for that).
If you have two "big" nodes with drbd and an third small one, you have all you need for HA.

Udo

 

[dietmar]

This is the reason why I wanted to know each nodes role in a 3-node HA cluster.

All nodes have the same 'role' (We do not have/need a dedicated management node).