USHIFT-197: Enable resource accounting in MicroShift systemd service

[oglok]

Signed-off-by: Ricardo Noriega rnoriega@redhat.com

Closes #USHIFT-197

[oglok]

With this PR, local testing showed no error message:

I0718 15:57:18.377000 2180849 server.go:662] "Failed to get the kubelet's cgroup. Kubelet system container metrics may be missing." err="cpu and memory cgroup hierarchy not unified.  cpu: /user.slice, memory: \
/user.slice/user-1000.slice/session-1.scope" 

With these options in the systemd service file, we enable accounting as you can see in the status of the service itself:

systemctl status microshift
● microshift.service - MicroShift
   Loaded: loaded (/usr/lib/systemd/system/microshift.service; enabled; vendor preset: disabled)
   Active: active (running) since Wed 2022-08-17 11:01:57 EDT; 8min ago
 Main PID: 1821 (microshift)
    Tasks: 25 (limit: 14213)
   Memory: 464.5M
      CPU: 51.887s
   CGroup: /system.slice/microshift.service
           └─1821 /usr/bin/microshift run

[root@localhost ~]# systemctl show microshift | grep -i accounting
CPUAccounting=yes
IOAccounting=no
BlockIOAccounting=yes
MemoryAccounting=yes
TasksAccounting=yes
IPAccounting=no

[oglok]

If required, I could add a new target to the Makefile so developers can run MicroShift from systemd in an easier way as @mrunalp suggested.

[ggiguash]

Why are we resetting these 3 specific limits? How do we know we should not bump more to avoid the mentioned performance problems.

[ggiguash]

Also, can we add a link in the comment that describes the performance problem?

[oglok]

I'm not sure I follow @ggiguash , are you talking about the performance issues caused by a very high value in LimitNOFILE? I could reduce it to 1048576 as it seems to be the standard for other project.

[oglok]

coreos/fedora-coreos-tracker#329

[oglok]

LimitNOFILE = 1048576 seems to be the standard number for most of the projects that deal with containers from a systemd service.

[ggiguash]

Can we update the comment to something more concrete? I think it's important to note what part of the kernel is not performing well.

Having higher limit values cause file system performance problems due to accounting overhead in the kernel. 
We recommend using cgroup to do container-local accounting.

[openshift-ci]

@oglok: all tests passed!

Full PR test history. Your PR dashboard.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[ggiguash]

/lgtm

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ggiguash, oglok

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:

• OWNERS [ggiguash,oglok]

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

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ggiguash, oglok

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:

• OWNERS [ggiguash,oglok]

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