Zabbix monitoring - cluster

[bpbp]

Hi, I am using the official Zabbix template:
"Proxmox VE by HTTP"

I set up my first host and to my surprise it discovered all the nodes in my cluster and added items & triggers for them, but under the first host pve01. I think if I add the other 4 nodes in the cluster I am going to get a lot of duplicate items and triggers.

What is everyone else doing here? Just monitoring a single node?

 

[UdoB]

I think if I add the other 4 nodes in the cluster I am going to get a lot of duplicate items and triggers.

Yes, unfortunately I am experiencing the same.
To reduce the number of active triggers I have simply "Suppress Indefinitely" a lot of triggers.

My impression is that this is by design.

Best regards

 

[bpbp]

Thank you for confirming!
After sleeping on it I have decided to monitor each node via zabbix-agent independently - I will now check which Proxmox items from the official template that could be useful

 

[signifywinter]

I am coming to the same conclusion as bpbp. I'd like to use the HTTP API, but this doesn't really organize hosts and items the way I would prefer.

 

[garbled]

Yeah, I'm in the same boat here. I set it up on my single-node pve, just to see how it worked, before adding my large cluster in. I can't figure out a sane way to monitor the cluster without having duplicates of everything.

I mean I could throw proxmox behind an haproxy instance, but.. ugh.. that's alot of work just for monitoring.

I guess it's slightly moot because my zabbix server is in that cluster, but if one node drops and it happens to be the one I pull monitoring from all I'll see is the API alert, which, I suppose, is something. (Aside from the zabbix agent alerts for the actual linux agent on the nodes)

 

[tuxillo]

Is this still the case?

 

[joshdinsdale]

Just ran into this same issue myself after bringing a small 3-node cluster live in our environment. My solution, while not ideal, seems to be the least worst option for the time being.
What I've done is setup a 4th node entry in zabbix to represent the cluster. Ive setup an A record in DNS that lists the IP of each of the 3 nodes and added this to the 4th node in zabbix. This way we get a single hosts that can sort of round-robin to the API via whichever host.
I've then installed the zabbic agent to each node to do basic linux level monitoring using the linux template.

I think we might change this to use HAProxy eventually, this will deal with the issue of a node possibly not being available.

 

[schoeppi]

I' ve configured a reverse proxy in front of the proxmox API and let the temlate connect only cia this reverse proxy. In the proxy all proxmox nodes are configured as backends. This has also the advantage that monitoring the cluster is working if one ore more of the proxmox cluster are down.

Additionaly I am monitoring each node with zabbix agent to get the infos I need and which are not available via the template.