15
Spreading the load on a Proxmox cluster
I’m on the pathway for setting up a Proxmox cluster with 3 thin clients, 2 of which are deployed (but not the cluster until I get the third machine running). I’m thinking of installing 4 VMs as follows: a) one for OpenWRT so that I can get rid of my SKY router b) one for Opensense c) one for Frigate d) one for Homeassistant/OpenHabian e) one for everything else (about 8 services with docker) The thin clients are Optiplexes with 16-32 GB RAM. Do you think I am overdoing this? Basically I want to keep my network VMs free from everything else so that I don’t have to reboot these when I have to reboot the VM with all the services.
Comments
Without knowing what exactly you’re planning, OpenWRT and OPNSense seem redundant. Both are router/firewall, with the latter being much more capable. I can’t imagine it can’t take on both roles you’re planning for them, unless you need some support specific plugin?
Openwrt does some funky thing with SKY fiber (sky uses ipv6 to customer routers and their hub does the ipv6 to ipv4 translation), which i understand openwrt handles opensense doesn’t. I read it on a forum somewhere but still need to do my homework. Opensense does handle this particular ipv6 to ipv4 translation which i understand is done with some protocol not supported by opensense. Basically, its so i can get rid of the sky router on my network.
Kinda hot title tbh…
SPREAD THE LOAD FRODO!
What specific models are you running? I have a trio of Optiplexes on 7th and 8th gen Intel hardware just for Proxmox. It’s great, I can run pretty much whatever I want.
OPNSense already does everything OpenWRT can do, and more.
Frigate runs in Docker. You don’t need a separate VM for it.
I’ve heard of OpenHabian, but there seems to be a lot of tribalism around it.
A single VM can run much more than 8 services without so much as a sneeze.
That is actually part of the reason Docker exists; individual services can be restarted very quickly. I have 4 VMs dedicated to just Docker services, and rarely have to reboot an entire VM unless I specifically did something to it (maybe 3 or 4 times over the last year).
You won’t notice the downtime.
Are you sure? I’ve never seen a thin client with more resources than just enough to run the remote software. Do you mean micro PCs?
Not sure why you need two routers, either. But it seems reasonable if they have enough resources to run all your stuff on two nodes, because you’ll be able to live migrate them around to reboot the nodes.
Note that if you pass through any PCI devices, like a GPU for frigate, you won’t be able to migrate or suspend that VM.
Are you trying to do HA with the network VMs? If so, maybe check that failover for the network devices works the way you expect.
Are you overdoing it? Probably. But most of us are. That’s part of the fun.
Yes, HA for network
You don’t need that many VMs. For networking get a dedicated box (can be a used consumer device) and then create a few VMs to run podman containers.
If you do want a dedicated instance for a service use LXC
The reason i’m not considering a dedicate box for Openwrt/opesense is due to the redundncy offered by the proxmox cluster which i would loose with a dedicated machine.
I’m not sure you understand how that works
The proxmox cluster still need a network.
Sign in to comment. Sign in