Managing Multiple Nodes with Different Hardware Config: Cluster or Datacenter Manager?

[SInisterPisces]

Hello,

I've got a couple of Proxmox nodes that aren't connected in any way yet. They're both Intel 12th Gen CPU-based: one is a 12700T mini PC, and the other is an N100 SBC. Their hardware is very much not identical, but use the same CPU family (though the 12700T has a more powerful iGPU).

I'd like to manage them from a single interface. I was planning to set up a cluster (I realize I'd need to be careful with HA-based live migration due to the disparate hardware on each node), but at this point I'm wondering if the Proxmox Datacenter Manager might be the better choice given that they're not identical nodes.

One thing I was looking forward to when I clustered them was being able to define backup jobs and external storage and SDN configurations in one place for both nodes without having to keep track of identical configs on two separate nodes--I'm not sure those are things Datacenter Manager can do yet?

So unless there's a really good reason to avoid it, I think a cluster is still be the best way to go for unified management?
Note: I'm was planning to set up a logging server and RAM-based logging on my nodes to try to mitigate the additional boot disk I/O that clustering introduces, and also run a separate q-device, so I'd already planned on mitigating two of the bigger pain points with clustering. PDM wasn't a thing when I started considering how to set this up.

I'd appreciate any advice. Thanks!

 

[Hend2002]

Hey!

I'm sorry you never got a response. I have the exact same problem right now.

I'd really appreciate it if you could share how this worked out for you.
How did you end up managing the nodes? Did you set up a cluster, use the Datacenter Manager, or try both? And what made settle on that?

Thanks!

 

[SInisterPisces]

Life intervened and danced on my head for a while, so I ended up having to get rid of the second PVE node I had. I'm also planning to get rid of the mini PC I have now and centralize my main production workloads into a 12th gen Intel Dell Precision 3660. So, I learned enough to realize I had too much hardware and needed to start over. Oops.

In the meantime, I've re-evaluated my strategy a bit. As much as I'd like to set up a real cluster and have live migration and centralized management inside the PVE GUI itself, the more I dug into it the more I realized how much more management burden it introduces:

1. I was always more interested in centralized management than in live migration or HA.
2. You've got to have a quorum, so that means having to manage a q-device if you've only got two PVE nodes--and if you've got three, you can skip the q-device but have to be mindful of when and how devices reboot or shut down, or what happens if something loses power and takes your cluster down with it.
3. Enough people were recommending a standalone switch for the cluster network even in a small/homelab environment that I realized that's effectively required if I wanted to be running anything close to a recommended configuration that I could get help with if I had trouble. (Otherwise, if I had problems, a lot of initial advice was going to be "get a dedicated switch.")
4. I'm relatively confident the Sabrent Rocket 4/Rocket 4 Plus NVMEs in my machines would hae handled the write-load from the cluster logs, especially since I was going to offload that to a NAS, but again, that was another layer of complexity.
5. I'm behind on actually using my PVE setup to do fun stuff because I keep going down infrastructure rabbit-holes. I'm learning a lot, but I'm not having the most fun.
6. I'm currently planning three separate Proxmox nodes with three very different hardware configurations; the clustering system is, I think, most useful with (nearly-)identical nodes.

All that said, while I still think setting up a cluster would be fun and I'd like to do it some day, I've realized it's overkill for my needs, and I need to actually start using my stuff and enjoying it.

All that to say that I'm effectively looking at running at least three single-node "clusters" that don't really need to talk to each other, which means Datacenter Manager should be the best choice, even if it's not feature-complete yet. From what I can see, it has all the features I currently need. It's also a simpler approach that lets me focus more on using my setup, and less on endlessly tinkering with it.

Right now, my current plan is this a TrueNAS server combined with three single-node Proxmox clusters managed by PDM:

1. Network Storage: I have a TrueNAS system set up for persistent storage including local backups, and I'm going to be testing the new NVME-oF-based shared storage technology for my three "clusters."
2. License Management/Enterprise Repo Access: I have some community licenses mostly because I wanted to support the project. PDM can manage those licenses and coordinate sharing them between each node that it manages (this is a newer feature, and very welcome versus having to go log into the Proxmox store to check my license status).
3. Andromeda I - Main productivity workload:
   1. Dell Precision 3660T (12900K, 128 GB DDR5 (from the before times), NVIDIA A2000 6 GB, 4x10 Gbps NICs). With its specs, I can pretty much do everything I want to do on it as far as actual production work, tools, games, and experiments.
   2. Meant to be always on, but won't break my network if it's turned off.
4. Cassiopeia I - Proxmox Backup Server, Network Logging, Network Metrics:N300-based system with 24-48 GiB of RAM (depending on how I distribute the sticks I've got), 1x10 Gbps NIC, and a bunch of scrounged enterprise SATA SSDs.
   1. Right now, my only PBS node is a Tuxis server in the Netherlands. I love the service, but I want a local PBS again.
   2. My TrueNAS is my main backup machine, and the local PBS server will back up to the TrueNAS, but I want a dedicated box that only handles Proxmox backup and logging for Proxmox and the rest of the network.
   3. My view here is that these sorts of workloads should be set it and forget it type things--maybe you log in to read your logs, but it should just sit there and not be messed with, so it stays stable. It's also my lowest power machine for the amount of performance it has, so I don't mind having it on 24/7.
   4. Most importantly, it should stay up and undisturbed if anything else goes down, and hopefully log whatever blew up and why.
5. Pisces I - Network Operations Management: Intel 8th Gen NUC with 16 GiB of ancient DDR4 laptop RAM, 1 TB internal storage for VMs, a small enterprise SATA SSD for the boot drive, and 1x 10 Gbps NIC.
   1. Control VM/LXC containers for, e.g., Unifi, the other two Proxmox clusters.
   2. Proxmox Datacenter Manager will live here.
   3. Lightweight Ubuntu Server with IceWM for interacting with web GUIs for Proxmox VE, Proxmox Datacenter Manager, TrueNAS, monitoring and logging GUIs, etc. I'll have a Firefox install set up with defined tab layouts so I have a very focused environment that just lets me see and manage all my stuff with no distractions.
   4. Won't break anything when the system is off.

That's the plan, at least. Ask me again in a few months how it went.