Uploaded image for project: 'Red Hat Advanced Cluster Management'
  1. Red Hat Advanced Cluster Management
  2. ACM-27906

ACM: Assisted install fails in disconnected env due to Container image availability check pulls from upstream quay.io despite ClusterImageSet pointing to local mirror

XMLWordPrintable

    • False
    • Hide

      None

      Show
      None
    • False
    • Important
    • None

      Container image availability check pulls from upstream quay.io instead of mirrored registry in disconnected environment,

      despite ClusterImageSet pointing to local mirror .

      Description of problem:
      in a disconnected environment, the ClusterImageSet (CIS) is correctly configured to point to the local mirrored registry.
      However, during the container-image-availability validation step on the spoke node, the assisted-installer agent attempts to pull the OCP release image directly from quay.io instead of using the mirrored image specified in the ClusterImageSet. This causes the installation to get stuck in "preparing-for-installation" state indefinitely.

      ClusterImageSet is correctly configured to use mirrored image:

      changed: [registry.kni-qe-59.lab.eng.rdu2.redhat.com] => {
        "cmd": "oc get clusterimageset 4.20 -o json | jq -r .spec.releaseImage\n",
        "stdout": "registry.kni-qe-59.lab.eng.rdu2.redhat.com:5000/openshift-release-dev/ocp-release:4.20.8-x86_64"
      }

      container-image-availability check uses upstream quay.io URL instead of mirrored URL:

      oc -n multicluster-engine logs -l app=assisted-service
      time="2025-12-22T16:02:43Z" level=info msg="Submitting step <container-image-availability> ... Arguments: <[{\"images\":[
        \"quay.io/openshift-release-dev/ocp-release:4.20.8-x86_64\",   <-- WRONG: Using upstream URL
        \"quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:fa135c50d22ae6254b9c6bdfc492e156c757ad82c4516bf5bc503a4c954f26b8\",
        \"quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:595874ce69fb9eae354d8ab748d06f5997f2fe9070c29adee6ae3dbdbb0fa37d\",
        \"registry.redhat.io/multicluster-engine/assisted-installer-rhel9@sha256:37355649b26095d8b94a72638c3759ef8a65e3f501e2b3a72df705cee9018a6a\"
      ],\"timeout\":960}]>"

       

      Version-Release number of selected component (if applicable):
      HUB OCP VERSION 4.21.0-ec.3
      SPOKE OCP VERSION 4.20.8
      ACM 2.15.1
      MCE 2.10.1

      How reproducible:

      often

      Steps to Reproduce:

      1. Set up a disconnected hub cluster with a local mirror registry
      2. Mirror the OCP release images to the local registry
      3. Create a ClusterImageSet pointing to the mirrored image
      apiVersion: hive.openshift.io/v1
      kind: ClusterImageSet
      metadata:
        name: "4.20"
      spec:
        releaseImage: registry.kni-qe-59.lab.eng.rdu2.redhat.com:5000/openshift-release-dev/ocp-release:4.20.8-x86_64 
      1. Deploy a spoke cluster using ZTP/GitOps with AgentClusterInstall referencing the ClusterImageSet .

       

      Actual results:
      The container-image-availability check attempts to pull from the upstream quay.io/openshift-release-dev/ocp-release:4.20.8-x86_64 URL, which fails in the disconnected environment, causing the installation to hang.

      Expected results:

      The container-image-availability check should use the mirrored image URL from the ClusterImageSet (registry.kni-qe-59.lab.eng.rdu2.redhat.com:5000/openshift-release-dev/ocp-release:4.20.8-x86_64) when validating image availability on the spoke node.

      Additional info:

        1. journal-collect.log
          4.65 MB
          Bahaa Azem
        2. journal-collect5.log
          2.47 MB
          Bahaa Azem

              eerez@redhat.com Elior Erez
              rh-ee-bazem Bahaa Azem
              Vladislav Kolodny Vladislav Kolodny
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated: