[SOLVED] Node configuration

[kenyoukenme]

Hi all. I wanna ask a few questions:

1. if proxmox nodes in HA cluster must have the same specs (ram, cores). Ive read the documentation for HA but im not sure if it is implied that nodes must have the same specs. i will create a test cluster anyways this week, but would save me time if i have answer for this question.

2. And if its okay for nodes not to have the same specs, is there any risk/downside.

3. my boss wants to save money (hence proxmox) but our cloud provider doesnt give us much flexibility regarding the specs of the cloud and the most cost-effective plan for us is to have 4 dedicated server (4 nodes). but as we all know even nodes wont work, and to get around this i can have a 3 voting HA cluster node and 1 voting non-HA cluster member node. is there any documentation or related links for this setup?

im using Proxmox 5.4 on Debian 9.

thanks guys in advance

 

[t.lamprecht]

1. if proxmox nodes in HA cluster must have the same specs (ram, cores). Ive read the documentation for HA but im not sure if it is implied that nodes must have the same specs. i will create a test cluster anyways this week, but would save me time if i have answer for this question.

First, there are no checks whatsoever which would disallow a setup if one node is a glorious supercomputer and the other a potato, as long as clustering (corosync and it's network) between them works the HA stack won't complain.
That said, the reason for why we say this: if you use HA, all HA services from a node must be able to run on the remaining ones if an outage of a node happens. So if you'd use one very powerful server with >1TB memory and 4way NUMA with hundreds of CPU core hosting a few hundred HA managed VMs, and two other server with just a few GB memory and a few cores, then the outage of the big server will overload and kill the other two servers => no HA.

Thus, while the specs do not need to be identical, they should be in the same order of magnitude and be dimensioned with a reserver CPU, Memory (and maybe IO) load for handling the additional HA services it may need to host during an node failure.

2. And if its okay for nodes not to have the same specs, is there any risk/downside.

see above

3. my boss wants to save money (hence proxmox) but our cloud provider doesnt give us much flexibility regarding the specs of the cloud and the most cost-effective plan for us is to have 4 dedicated server (4 nodes). but as we all know even nodes wont work, and to get around this i can have a 3 voting HA cluster node and 1 voting non-HA cluster member node. is there any documentation or related links for this setup?

Even nodes work just fine, the thing is that even if you use 4 nodes, still only one can fail without impacting the whole cluster (and it's quorum) - thus, you did not really win anything in comparission to the three--node cluster in the "how many node can fail at the same time" regard. So, you could use them all for HA just fine. If you still do not want, then just create a HA group which has the 4th node with lower priority in there and assign all HA services to that group.

In general, you may want to take a look at our docs:

https://pve.proxmox.com/pve-docs/chapter-ha-manager.html

An additional, external, cluster vote could be also for interest of you: https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support

 

[kenyoukenme]

thanks alot mate!

Indeed four nodes or any number of nodes work just fine.
One must read the corosync votequorum manual haha