CPU utilisation graph on docker containers

CMK version:2.1.0p16
OS version:checkmk/check-mk-raw:2.1.0p16

Error message:
Hi,

I noticed on docker containers, monitored with mk_docker.py or mk_docker2.py that if you view the the CPU utilisation graph the percentage is locked to 100% but values are sometimes 400% or 600% on multicore systems.
I think the graph should scale up or values should be calculated compared to all cores like on non docker containers if possible. For example on bare metal multicore systems usage never goes over 100% on multicore systems, values are calculated based on all cores.

Below images show the bare metal system on wich the docker container is also running. The bare metal system has 64Cores and the container is limited to 16Cores via compose.

Is this Normal behavior ?

Thanks.

Here are images for bare metal and docker container

1 Like

I have a similar problem, not on docker container but on LXC container. The container has 4 CPUs, and the graph is not useful because the utilisation is always over 100%. It seems the number of CPUs is not taken into consideration for checkmk agent on containers?

1 Like

I’m experiencing the same problem on 2.1.0p21.cme.

It seems to me, that there is no dedicated check function for the docker_container_cpu check but instead a “SectionCpuUtilizationOs” object is returned which apparently handles the checking (see checkmk/docker_container_cpu.py at 158833c70eca216170cdd4b16c2955106f5ce14e · tribe29/checkmk · GitHub).

Maybe someone with the some insight into the docker plugin can shed some light on this.

I just found this Werk in the upcoming release 2.1.0p27: cpu_utilization_os: Allow displaying more than 100% in service graph

This should fix the issue.

1 Like

How can I be notified when 2.1.0p27 is ready? Is there something like a mailing list for release announcement?

you can subscribe to announcements in this forum or on lists.mathias-kettner.de Mailing Lists

2 Likes

This topic is temporarily closed for at least 4 hours due to a large number of community flags.

This topic was automatically opened after 4 hours.