nmstate metrics pod unhealthy. Keeps crashing.

[kristvanbesien]

I deployed the kubernetes-nmstack operator, following the instructions listed at:
https://github.com/nmstate/kubernetes-nmstate/releases/tag/v0.85.1

Not everything appears to deploy fine:

% kubectl get pods -n nmstate
NAME                                    READY   STATUS             RESTARTS       AGE
nmstate-cert-manager-745bf8f4fc-vhrsr   1/1     Running            0              12m
nmstate-handler-64kb2                   1/1     Running            0              12m
nmstate-handler-lhc8b                   1/1     Running            0              12m
nmstate-handler-zgjng                   1/1     Running            0              12m
nmstate-metrics-74ddd9c697-47wvd        1/2     CrashLoopBackOff   5 (118s ago)   5m21s
nmstate-operator-68dcbf8c55-vfdzt       1/1     Running            0              12m
nmstate-webhook-58cc4b5957-tnds5        1/1     Running            0              12m

Further investigation turned up that the problem is that the metrics pod contains a container that uses an image from origin:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nmstate-metrics
  namespace: nmstate
 (...)
spec:
  replicas: 1
  selector:
    matchLabels:
      name: nmstate-metrics
  template:
(...)
    spec:
      containers:
(...)
      - name: kube-rbac-proxy
        image: quay.io/openshift/origin-kube-rbac-proxy:4.10.0
(...)

And here is the problem: Images from Origin are only good for the intel architecture, and my cluster runs on arm...

I manged to fix this by changing the value of the KUBE_RBAC_PROXY_IMAGE in the operator.yaml, to quay.io/brancz/kube-rbac-proxy:v0.20.1

So please consider for the next release to make sure that all image references point to multi arch images.

[qinqon]

Thanks, we will look into this, in fact I think with newere controller-runtime versions we don't heed the kube rba proxy image