Uploaded image for project: 'OpenShift Virtualization'
  1. OpenShift Virtualization
  2. CNV-80055

OLM bundle unpack fails with DeadlineExceeded for HCO subscription; InstallPlan never created

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Critical Critical
    • None
    • CNV v4.18.30
    • CNV CI and Release
    • None
    • Quality / Stability / Reliability
    • 0.42
    • False
    • Hide

      None

      Show
      None
    • False
    • None
    • CNV QE DevOps Sprint 284
    • None

      Summary

      When deploying CNV 4.18 via verify-cnv-4.18.z-build (or deploy_cnv.sh) using a custom CatalogSource with IIB image brew.registry.redhat.io/rh-osbs/iib:1106018, OLM's bundle unpack job exceeds its deadline and fails with DeadlineExceeded. The subscription never gets an InstallPlan, so HCO/CNV is never installed.

      The CatalogSource is healthy (READY). The failure is in OLM's bundle unpack step, not in the catalog or our script timeout.

      Steps to reproduce

      1. Provision or use a cluster (e.g. OCP 4.18 on RHOS-PSI). Example: infd-vrf-418t0 (rhos-psi.cnv-qe.rhood.us).
      2. Run verify-cnv-4.18.z-build (or equivalent deploy that creates a CatalogSource from the IIB and then a Subscription to kubevirt-hyperconverged).
      3. Job parameters used: IIB_IMAGE=registry-proxy.engineering.redhat.com/rh-osbs/iib:1106018 (or brew.registry.redhat.io/rh-osbs/iib:1106018), HCO_REGISTRY_IMAGE from Konflux, channel=nightly, startingCSV=kubevirt-hyperconverged-operator.v4.18.30.
      4. Wait for the "Deploy CNV-4 build" stage. The script waits for the catalog to be READY (succeeds), applies the subscription, then polls for status.installPlan.name on the subscription.
      5. Observe: InstallPlan never appears; job fails with "Timeout waiting for InstallPlan of HCO operator to be created" (script timeout 100s on cnv-4.18).
      6. On the cluster, inspect the subscription and catalog-operator logs (see Evidence below).

      Expected result

      • OLM unpacks the bundle from the catalog and creates an InstallPlan for the subscription.
      • Our script (or manual approval) can then approve the InstallPlan and HCO install proceeds.

      Actual result

      • OLM starts unpacking the bundle but the unpack Job exceeds OLM's internal deadline.
      • Subscription status shows: BundleUnpackFailed with message "bundle unpacking failed. Reason: DeadlineExceeded, and Message: Job was active longer than specified deadline".
      • No InstallPlan is ever created (subscription.status.installPlan remains empty).
      • CatalogSource remains READY and healthy; the catalog pod serves the registry successfully.

      Evidence

      Jenkins:

      Cluster (infd-vrf-418t0):

      Subscription conditions (openshift-cnv/hco-operatorhub):

      CatalogSourcesUnhealthy: AllCatalogSourcesHealthy - all available catalogsources are healthy
      BundleUnpacking: UnpackingInProgress -
      BundleUnpackFailed: BundleUnpackFailed - bundle unpacking failed. Reason: DeadlineExceeded, and Message: Job was active longer than specified deadline
      

      Catalog-operator log (openshift-operator-lifecycle-manager):

      time="2026-02-17T05:29:25Z" level=info msg="checking for existing installplan" channel=nightly id=5GkF1 namespace=openshift-cnv pkg=kubevirt-hyperconverged source=hco-catalogsource sub=hco-operatorhub
      time="2026-02-17T05:29:25Z" level=info msg="unpacking bundles" id=5GkF1 namespace=openshift-cnv
      time="2026-02-17T05:29:25Z" level=info msg="bundle unpacking failed. Reason: DeadlineExceeded, and Message: Job was active longer than specified deadline" id=5GkF1 namespace=openshift-cnv
      

      CatalogSource (openshift-marketplace/hco-catalogsource):

      • spec.image: brew.registry.redhat.io/rh-osbs/iib:1106018
      • status.connectionState.lastObservedState: READY
      • Pod hco-catalogsource-* is 1/1 Running.

      Subscription spec: source: hco-catalogsource, sourceNamespace: openshift-marketplace, channel: nightly, startingCSV: kubevirt-hyperconverged-operator.v4.18.30. No status.installPlan.

      Impact

      • verify-cnv-4.18.z-build (and any deploy using this IIB/catalog) fails to install CNV on the cluster.
      • Workaround: increase OLM's bundle unpack job deadline (if configurable) or use a smaller/faster-to-unpack catalog. Our automation has also increased the script-side wait for InstallPlan (5 min) so that when unpack does succeed, the job does not time out prematurely.

      Environment

      • Cluster: infd-vrf-418t0.rhos-psi.cnv-qe.rhood.us (OCP 4.18)
      • IIB: brew.registry.redhat.io/rh-osbs/iib:1106018 (~1.9GB image)
      • Branch: cnv-qe-automation cnv-4.18 (verify job uses CNV_QE_AUTO_BRANCH=cnv-4.18)

              lbednar@redhat.com Lukas Bednar
              lbednar@redhat.com Lukas Bednar
              Albert Jubany Juarez
              Daniel Keler Daniel Keler
              Votes:
              0 Vote for this issue
              Watchers:
              5 Start watching this issue

                Created:
                Updated: