ceph monitoring via zabbix

[wylde]

http://docs.ceph.com/docs/mimic/mgr/zabbix/

I'm curious if anyone has tried this? I'm experiencing an oddity I believe. I'll install this on all nodes which are part of the ceph cluster, but every time I set the identifier it replicates throughout all of the nodes. For example

install on node-1 and set the identifier to node-1
install on node-2 and set the identifier to node-2
now if you go back to node-1 and run 'ceph zabbix config-show' you'll notice that its identifier is set to node-2

This effectively limits the monitoring source to a single node as zabbix will refuse incoming data from a source which isn't correctly identified.

 

[wylde]

I was incorrect and have since found out how this actually works. Ceph elects a manager process which does all of the metric sending. So in zabbix, you create a generic host with just this template assigned to it and this will allow for the various managers as they get elected to send data into zabbix land.

Hopefully, this will help others