Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-3029

BareMetalHost CR fails to delete on cluster cleanup

XMLWordPrintable

    • Important
    • None
    • 5
    • Metal Platform 227, Metal Platform 228, Metal Platform 229, Metal Platform 230, Metal Platform 232, Metal Platform 233
    • 6
    • Rejected
    • False
    • Hide

      None

      Show
      None
    • Hide
      Before this update, when an attempt is made to delete a namespace that contains baremetalhost custom resource definitions, deletion may hang indefinitely due to inability of the baremetalhost finalizer to complete successfully. This is because the finalizer depends on creation of preprovisioning image which can’t succeed because the namespace is already being deleted. Currently, this behavior is altered and deletion of the namespace is no longer blocked indefinitely by the baremetalhost finalizer.

      https://issues.redhat.com/browse/OCPBUGS-3029
      Show
      Before this update, when an attempt is made to delete a namespace that contains baremetalhost custom resource definitions, deletion may hang indefinitely due to inability of the baremetalhost finalizer to complete successfully. This is because the finalizer depends on creation of preprovisioning image which can’t succeed because the namespace is already being deleted. Currently, this behavior is altered and deletion of the namespace is no longer blocked indefinitely by the baremetalhost finalizer. https://issues.redhat.com/browse/OCPBUGS-3029
    • Bug Fix
    • Proposed
    • Customer Escalated
    • Hide
      1/3: Moving this to 4.12 POSTGA for Telco, not impacting Nokia (they use Billi, the impact here is on ZTP with ACM).  Needs a release note for 4.12 GA.
      12/19: Yellow. Waiting on baremetal team to determine how to disable autocleaning on baremetal host
      12/15: pending next steps based on Telco findings re: ArgoCD
      12/12: Telco Engineering investigating re: possible to control the cleanup flow with ArgoCD
      12/8: Waiting on Telco to test various workarounds and answer questions.
      11/30: changed Telco rank/bucket to 2 since this currently looks like it'd be more of a lab issue (albeit w/ the potential to become a production thing, and albeit the availability of an ugly workaround)
      11/29: Waiting on bug owner to analyze data provide by Telco and debug live
      11/28: Waiting on info from Telco
      11/22: Waiting on info from Telco
      11/21: Y e l l o w. Triage is in progress.
      11/9: proposing this be a Release Blocker for 4.12 since it impacts installation
      11/4: new to the Telco-Grade OCP 4.12 gating list
      Rel Note for Telco: Yes, proposed release note text added by Ian M in this jira
      Show
      1/3: Moving this to 4.12 POSTGA for Telco, not impacting Nokia (they use Billi, the impact here is on ZTP with ACM).  Needs a release note for 4.12 GA. 12/19: Yellow. Waiting on baremetal team to determine how to disable autocleaning on baremetal host 12/15: pending next steps based on Telco findings re: ArgoCD 12/12: Telco Engineering investigating re: possible to control the cleanup flow with ArgoCD 12/8: Waiting on Telco to test various workarounds and answer questions. 11/30: changed Telco rank/bucket to 2 since this currently looks like it'd be more of a lab issue (albeit w/ the potential to become a production thing, and albeit the availability of an ugly workaround) 11/29: Waiting on bug owner to analyze data provide by Telco and debug live 11/28: Waiting on info from Telco 11/22: Waiting on info from Telco 11/21: Y e l l o w. Triage is in progress. 11/9: proposing this be a Release Blocker for 4.12 since it impacts installation 11/4: new to the Telco-Grade OCP 4.12 gating list Rel Note for Telco: Yes, proposed release note text added by Ian M in this jira

      Description of problem:

      OCP cluster installation (SNO) using assisted installer running on ACM hub cluster. 
      Hub cluster is OCP 4.10.33
      ACM is 2.5.4
      
      When a cluster fails to install we remove the installation CRs and cluster namespace from the hub cluster (to eventually redeploy). The termination of the namespace hangs indefinitely (14+ hours) with finalizers remaining. 
      
      To resolve the hang we can remove the finalizers by editing both the secret pointed to by BareMetalHost .spec.bmc.credentialsName and BareMetalHost CR. When these finalizers are removed the namespace termination completes within a few seconds.

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

      OCP 4.10.33
      ACM 2.5.4

      How reproducible:

      Always

      Steps to Reproduce:

      1. Generate installation CRs (AgentClusterInstall, BMH, ClusterDeployment, InfraEnv, NMStateConfig, ...) with an invalid configuration parameter. Two scenarios validated to hit this issue:
        a. Invalid rootDeviceHint in BareMetalHost CR
        b. Invalid credentials in the secret referenced by BareMetalHost.spec.bmc.credentialsName
      2. Apply installation CRs to hub cluster
      3. Wait for cluster installation to fail
      4. Remove cluster installation CRs and namespace

      Actual results:

      Cluster namespace remains in terminating state indefinitely:
      $ oc get ns cnfocto1
      NAME       STATUS        AGE    
      cnfocto1   Terminating   17h

      Expected results:

      Cluster namespace (and all installation CRs in it) are successfully removed.

      Additional info:

      The installation CRs are applied to and removed from the hub cluster using argocd. The CRs have the following waves applied to them which affects the creation order (lowest to highest) and removal order (highest to lowest):
      Namespace: 0
      AgentClusterInstall: 1
      ClusterDeployment: 1
      NMStateConfig: 1
      InfraEnv: 1
      BareMetalHost: 1
      HostFirmwareSettings: 1
      ConfigMap: 1 (extra manifests)
      ManagedCluster: 2
      KlusterletAddonConfig: 2

       

            janders@redhat.com Jacob Anders
            rhn-support-imiller Ian Miller
            Dmitry Dmitriev Dmitry Dmitriev
            Votes:
            0 Vote for this issue
            Watchers:
            23 Start watching this issue

              Created:
              Updated:
              Resolved: