I have an RKE1 cluster on SuSE Linux Enterprise VMs running on VMWare. The Checkmk agent (RPM) was installed on all machines that form the RKE cluster.
I assume, that metrics for this check “come” from the agent on the host and from the special agent, too, which is probably the reason why it says “found 2 times”.
How do I “solve” this problem, without simply “disabling” the “Kernel Performance” service and/or check? Can I somehow set “precedence” which metrics to prefer (e.g. “prefer agent metrics”)?
A workaround I applied is to skip the kernel section of the agent I installed on all hosts
via /etc/check_mk/exclude_sections.cfg where I added MK_SKIP_KERNEL="yes" on all
affected hosts.
This of course means, that I now have to rely an the special agent for this check.
I’d prefer to be able to keep the metrics from the “normal” and the special agent.
It would be great, if Checkmk would use any available information in such a scenario, or
if it could be configured somehow. For the time being, I’m OK again:
Force of habit: I had a bunch of hosts, I could access them, I installed agents on them…
I thought it’s nice to have a “fallback” if metrics can no longer be collected through kubernetes, for any reason.
Thanks for pointing out that I could disable the “machine sections” part in such a scenario. Perhaps this should be mentioned somehow in the official documentation.
the recommended solution would be to not install agents on these hosts.
The Kubernetes Collector automatically collects machine metrics (e.g. Filesystem, Network, CPU, Memory, Kernel) for each node in a cluster. If you e.g. add another node to a cluster, it is automatically monitored. If your cluster has auto-scaling functionality, then this becomes very important.
Thus, we do not recommend disabling the machine sections. Therefore, we also do not document it. Cheers!
