Monitor Linux and Ceph at the same time

[mrosu]

We have multiple hyperconverged Proxmox + CEPH servers and I am running into an issue monitoring them with the new Zabbix / CEPH Agent interface. We use Grafana to display the data, and we are pointing the CEPH graphs to a single Zabbix Host named cephcluster.

Previous Setup:

We were monitoring the Proxmox hosts using the Linux agent.

We pushed data to zabbix using the old (now deprecated) Zabbix Module. This was great, because we could push to the same Zabbix Host (called cephcluster, which is separate from node1, node2, etc) from multiple servers to add data redundancy in case one of the hosts was down. Note that we were using a totally separate Host for the Ceph Cluster than for the individual server Hosts (Linux monitoring).

I attached an diagram to help clarify the previous setup.








































New Setup:

We are monitoring the Proxmox hosts using the Linux agent, as before.

How are we supposed to use the Zabbix agent to monitor ceph with data redundancy? Is this possible?

Do we have no choice but to monitor the ceph cluster using each individual nodeX host and pull data to grafana from a single node host?

Thank you for reading this. I appreciate any assistance I can get.

[mrosu]

I am posting my findings and solution here:

The CEPH data is only available from the Active CEPH Manager, which is only 1 active at a time.

• Pull data from each Proxmox host and save it to that respective host: node1, node2, node3 etc.
• In Zabbix, create a Host Group: Proxmox Hosts or CEPH hosts, or whatever you want. All Proxmox/CEPH hosts need to be listed here.
• In Grafana, when pulling data, select Group: Proxmox/CEPH Group, select Host: /.*/

By doing this, it will pull data from whichever host has the live data available and ignore the other hosts.

This gives redundancy to viewing the data from the various hosts, even when the host manager changes from host to host during cluster maintenance or what have you.

[mrosu]

I am posting my findings and solution here:

The CEPH data is only available from the Active CEPH Manager, which is only 1 active at a time.

• Pull data from each Proxmox host and save it to that respective host: node1, node2, node3 etc.
• In Zabbix, create a Host Group: Proxmox Hosts or CEPH hosts, or whatever you want. All Proxmox/CEPH hosts need to be listed here.
• In Grafana, when pulling data, select Group: Proxmox/CEPH Group, select Host: /.*/

By doing this, it will pull data from whichever host has the live data available and ignore the other hosts.

This gives redundancy to viewing the data from the various hosts, even when the host manager changes from host to host during cluster maintenance or what have you.