The kubelet in Kubernetes creates a cgroup /docker-daemon and attempts to
move the docker daemon into that cgroup as part of its internal
accounting.  It also lowers the OOM score of the daemon to -900 so the
oom_killer will not target it in low memory situations over end-user
containers.

On systemd environments, this type of cgroup manipulation is not needed
because systemd already launched the daemon in the cgroup associated with
its service, i.e. /system.slice/docker.service.  I recommend modifying the
unit file for your docker daemon and setting the OOMScoreAdjust=-900 to
ensure that in low-memory situations the oom_killer does not incorrectly
target your docker daemon, but in general, the kubelet should not be doing
this type of bootstrap manipulation on systemd environments.

I have a PR upstream that attempts to remove this noise when bootstrapping
on systemd that hopefully will get merged soon:

https://github.com/kubernetes/kubernetes/pull/20748

Thanks,
Derek



On Thu, Jan 28, 2016 at 3:10 PM, Dean Peterson <[email protected]>
wrote:

> I keep seeing this in the openshift console.  Is this something to be
> concerned about?
>
> _______________________________________________
> users mailing list
> [email protected]
> http://lists.openshift.redhat.com/openshiftmm/listinfo/users
>
>
_______________________________________________
users mailing list
[email protected]
http://lists.openshift.redhat.com/openshiftmm/listinfo/users

Reply via email to