Bug 1835996

+++ This bug was initially created as a clone of Bug #1835112 +++

Description of problem:
When upgrading the zVM cluster from 4.2.29 to 4.3.18, it fails with the message, that the openshift cluster samples operator was not rolled out.
When viewing the log of the container, it shows the following error:


time="2020-05-13T06:27:05Z" level=info msg="watch event tests not part of operators inventory"
time="2020-05-13T06:28:33Z" level=info msg="Spec is valid because this operator has not processed a config yet"
time="2020-05-13T06:28:33Z" level=info msg="error reading in content : open /opt/openshift/operator/ocp-s390x: no such file or directory"
time="2020-05-13T06:28:33Z" level=info msg="CRDUPDATE file list err update"
time="2020-05-13T06:28:36Z" level=error msg="unable to sync: open /opt/openshift/operator/ocp-s390x: no such file or directory, requeuing"

When rsh into pod, it only has the directory /opt/openshift/operator/ocp-x86_64

Version-Release number of selected component (if applicable):
4.3.18

How reproducible:


Steps to Reproduce:
1. Start uograde from 4.2.29 to 4.3.18
2.
3.

Actual results:
Upgrade fails

Expected results:
Excpect upgrade to work

Additional info:

Deployment-YAML cluster-samples-operator:
kind: Deployment
apiVersion: apps/v1
metadata:
  name: cluster-samples-operator
  namespace: openshift-cluster-samples-operator
  selfLink: >-
    /apis/apps/v1/namespaces/openshift-cluster-samples-operator/deployments/cluster-samples-operator
  uid: c4a61f65-586c-11ea-b504-02462c000005
  resourceVersion: '46056999'
  generation: 6
  creationTimestamp: '2020-02-26T07:51:16Z'
  annotations:
    deployment.kubernetes.io/revision: '4'
spec:
  replicas: 1
  selector:
    matchLabels:
      name: cluster-samples-operator
  template:
    metadata:
      creationTimestamp: null
      labels:
        name: cluster-samples-operator
    spec:
      nodeSelector:
        node-role.kubernetes.io/master: ''
      restartPolicy: Always
      serviceAccountName: cluster-samples-operator
      schedulerName: default-scheduler
      terminationGracePeriodSeconds: 30
      securityContext: {}
      containers:
        - resources:
            requests:
              cpu: 10m
          terminationMessagePath: /dev/termination-log
          name: cluster-samples-operator
          command:
            - cluster-samples-operator
          env:
            - name: WATCH_NAMESPACE
              valueFrom:
                fieldRef:
                  apiVersion: v1
                  fieldPath: metadata.namespace
            - name: OPERATOR_NAME
              value: cluster-samples-operator
            - name: RELEASE_VERSION
              value: 4.3.18
            - name: IMAGE_JENKINS
              value: >-
                quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:7688ecdcb88ff3b29abf0180da08cd26e42d285151cb399c0e3af160c1b2305e
            - name: IMAGE_AGENT_NODEJS
              value: >-
                quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:50fa2dbf44ac1ab0487b6c69a2eb7f3513325a99dda91799af306c35a9f39ac4
            - name: IMAGE_AGENT_MAVEN
              value: >-
                quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:4c052eb17a15cb35c1253d4fbd849959047b06b478e628e4601609c1b25bd178
          ports:
            - name: metrics
              containerPort: 60000
              protocol: TCP
          imagePullPolicy: IfNotPresent
          volumeMounts:
            - name: samples-operator-tls
              mountPath: /etc/secrets
          terminationMessagePolicy: File
          image: >-
            quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:7b6920adfb6ca19d9105d292feac51867eba3ee46825c0a4d187024c4695e790
        - name: cluster-samples-operator-watch
          image: >-
            quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:7b6920adfb6ca19d9105d292feac51867eba3ee46825c0a4d187024c4695e790
          command:
            - cluster-samples-operator-watch
            - file-watcher-watchdog
          args:
            - '--namespace=openshift-cluster-samples-operator'
            - '--process-name=cluster-samples-operator'
            - '--termination-grace-period=30s'
            - '--files=/etc/secrets/tls.crt,/etc/secrets/tls.key'
          resources:
            requests:
              cpu: 10m
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: File
          imagePullPolicy: IfNotPresent
      serviceAccount: cluster-samples-operator
      volumes:
        - name: samples-operator-tls
          secret:
            secretName: samples-operator-tls
            defaultMode: 420
      dnsPolicy: ClusterFirst
      tolerations:
        - key: node-role.kubernetes.io/master
          operator: Exists
          effect: NoSchedule
        - key: node.kubernetes.io/unreachable
          operator: Exists
          effect: NoExecute
          tolerationSeconds: 120
        - key: node.kubernetes.io/not-ready
          operator: Exists
          effect: NoExecute
          tolerationSeconds: 120
      priorityClassName: system-cluster-critical
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 25%
      maxSurge: 25%
  revisionHistoryLimit: 10
  progressDeadlineSeconds: 600
status:
  observedGeneration: 6
  replicas: 1
  updatedReplicas: 1
  readyReplicas: 1
  availableReplicas: 1
  conditions:
    - type: Progressing
      status: 'True'
      lastUpdateTime: '2020-05-12T12:51:11Z'
      lastTransitionTime: '2020-02-26T07:51:16Z'
      reason: NewReplicaSetAvailable
      message: >-
        ReplicaSet "cluster-samples-operator-64fc49d87" has successfully
        progressed.
    - type: Available
      status: 'True'
      lastUpdateTime: '2020-05-12T14:05:30Z'
      lastTransitionTime: '2020-05-12T14:05:30Z'
      reason: MinimumReplicasAvailable
      message: Deployment has minimum availability.

--- Additional comment from gmontero on 2020-05-14 18:24:26 UTC ---

Yep there is an upgrade specific (vs. initial install) error path with s390/ppc64le that I now see based on the data provided with the bug,
which stems from bootstrapping as removed for those platforms, but having the payload imagestreams like tests and must-gather coming into
the samples operator.

Now, a couple of notes:
1) a process reminder, per the OCP process, I would need to fix this in 4.5 first, then 4.4.z, then 4.3.z; so it will take a bit
2) Jeremy Poulin and Renin Jose are in the process of trying to validate the s390 image in early 4.5 payloads to see if we can include
some samples finally ... not sure yet if/when that would move back to 4.4 and 4.3, but if it did, it would obviate the need for the fix.

I've cc:ed Jeremy on this bug and will send a needinfo to him for comment.

But in the interim, I'll start on 1)

--- Additional comment from wking on 2020-05-14 19:07:06 UTC ---

We're asking the following questions to evaluate whether or not this bug warrants blocking an upgrade edge from either the previous X.Y or X.Y.Z. The ultimate goal is to avoid delivering an update which introduces new risk or reduces cluster functionality in any way. Sample answers are provided to give more context and the UpgradeBlocker flag has been added to this bug. It will be removed if the assessment indicates that this should not block upgrade edges.

Who is impacted?  If we have to block upgrade edges based on this issue, which edges would need blocking?
  example: 100% of customers upgrading from 4.2 to 4.3 running s390/ppc64le.  Is there also an impact from 4.2 -> 4.2, 4.3 -> 4.4, etc.?
What is the impact?  Is it serious enough to warrant blocking edges?
  example: Samples sticks on arch-specific bug, CVO sticks on samples, update hangs.
How involved is remediation (even moderately serious impacts might be acceptable if they are easy to mitigate)?
  example: Clearing the attempted update resolves the issue.  There is no other remediation procedure.

--- Additional comment from gmontero on 2020-05-14 19:23:53 UTC ---

(In reply to W. Trevor King from comment #2)
> We're asking the following questions to evaluate whether or not this bug
> warrants blocking an upgrade edge from either the previous X.Y or X.Y.Z. The
> ultimate goal is to avoid delivering an update which introduces new risk or
> reduces cluster functionality in any way. Sample answers are provided to
> give more context and the UpgradeBlocker flag has been added to this bug. It
> will be removed if the assessment indicates that this should not block
> upgrade edges.
> 
> Who is impacted?  If we have to block upgrade edges based on this issue,
> which edges would need blocking?
>   example: 100% of customers upgrading from 4.2 to 4.3 running s390/ppc64le.

I believe this is 100% of customers upgrading from 4.2 to 4.3 running s390/ppc64le.

> Is there also an impact from 4.2 -> 4.2, 4.3 -> 4.4, etc.?
> What is the impact?  Is it serious enough to warrant blocking edges?
>   example: Samples sticks on arch-specific bug, CVO sticks on samples,
> update hangs.

I would expect these results on 4.2 -> 4.2, 4.3 -> 4.4.

We are currently working with the multiarch team to get samples vetted 
on s390 and perhaps ppc64le for 4.5.  But most likely that is several
weeks out for ppc64le and maybe a few days to a week for s390x.  So 
I would not expect this for 4.4 -> 4.5.

Then, a discussion on backporting content to 4.4 or 4.3 could occur, though
it is not a given that would be agreed upon.

> How involved is remediation (even moderately serious impacts might be
> acceptable if they are easy to mitigate)?
>   example: Clearing the attempted update resolves the issue.  There is no
> other remediation procedure.

running `oc delete configs.samples cluster` should reset the samples operator;
when it comes back up, it will treat things like an initial install and should 
bootstrap as removed, without misguided attempts to read non-existent content.

--- Additional comment from gmontero on 2020-05-14 19:36:28 UTC ---

Also, on my needinfo to Jeremy (though if anybody on cc: knows please feel free to chime in) - is there 4.3 -> 4.4 tests upgrade tests on s390x coming anytime soon?

--- Additional comment from gmontero on 2020-05-14 19:44:17 UTC ---

Lastly, to the originator and QA contact, can either of you reproduce the upgrade issue, and 
then run `oc delete configs.samples cluster` and observe the result.

Ultimately, an `oc get clusteroperator openshift-samples -o yaml` should confirm the reset worked and samples is available==true progressing==false degraded==false version set,
like you would get on an initial install.

--- Additional comment from bdonahue on 2020-05-14 20:19:48 UTC ---

   This would only affect s390. 4.2 was not released on ppc64.