Kubelet emits warning rather than exiting on invalid cgroup setup

[dashpole]

What type of PR is this?
/kind bug

What this PR does / why we need it:
This maintains the prior behavior of the kubelet when the cgroup setup is unsupported. It emits an error rather than exiting. This is important for those who run kubelets locally to test changes unrelated to monitoring or cgroup setup.

Which issue(s) this PR fixes:
Fixes #78950

Does this PR introduce a user-facing change?:

NONE

cc @mattjmcnaughton @roycaihw

[dashpole]

/sig node
/priority important-soon

[dashpole]

/retest

[roycaihw]

Thanks @dashpole. I can confirm this fixes #78950!

[mattjmcnaughton]

/lgtm

Thanks for this quick fix @dashpole !

[stealthybox]

I just ran into this in an LXC container running systemd where I've split up the cpu and cpuacct cgroup mountpoints.
I am struggling to get systemd to assign the unit to matching slices, so I can't run the kubelet service.

Here's more context on this error:
https://github.com/kubernetes/kubernetes/blob/d30fbab/pkg/kubelet/cm/container_manager_linux.go#L802-L815
It seems this was done intentionally.

How critical is this functionality?
Does this disable CPU/MemoryAccounting for all Pods?

[dashpole]

@stealthybox I responded in slack as well, but i'll repeat it here for anyone else who happens across this.
This specific functionality is only used for monitoring. If you rely on the kubelet or runtime system containers in the /stats/summary API and see the errors this PR is fixing or adding, then those metrics will be missing from the summary API.

[derekwaynecarr]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: dashpole, derekwaynecarr

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• cmd/kubelet/OWNERS [derekwaynecarr]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[stealthybox]

I can confirm I built this for arm64 and it solved my issue.

[stealthybox]

Please note that this also affects docker:

W0621 18:17:18.104998    1600 server.go:628] failed to get the kubelet's cgroup: cpu and memory cgroup hierarchy not unified.  cpu: /, memory: /user.slice.  Kubelet system container metrics may be missing.
W0621 18:17:18.105420    1600 server.go:635] failed to get the container runtime's cgroup: failed to get container name for docker process: cpu and memory cgroup hierarchy not unified.  cpu: /, memory: /system.slice/docker.service. Runtime system container metrics may be missing.
I0621 18:17:18.107885    1600 manager.go:146] cAdvisor running in container: "/sys/fs/cgroup/cpu"

[dashpole]

@stealthybox yes, this issue is independent of the container runtime used.