Uploaded image for project: 'OpenShift Core Networking'
  1. OpenShift Core Networking
  2. CORENET-5832

Impact statement request for OCPBUGS-49994 ovnkube-controller stuck in crashloopbackoff when upgrading to 4.18 with clusterNetwork set to multiple networks of the same IP family

XMLWordPrintable

    • Icon: Spike Spike
    • Resolution: Done
    • Icon: Critical Critical
    • None
    • None
    • None
    • False
    • None
    • False
    • If docs needed, set a value
    • Unset
    • ?
    • No
    • ?
    • False
    • ?
    • ?
    • ToDo
    • Untriaged
    • Not Supported
    • ---
    • ---
    • 0
    • Untriaged

      Impact statement for the OCPBUGS-49994 series:

      Which 4.y.z to 4.y'.z' updates increase vulnerability?

      • Any upgrade to 4.18.* will fail.

      Which types of clusters?

      There is no way to identify the cluster with PromQL, a cluster admin has to manually check.

      The cluster network configuration is with:

      • "OVNKubernetes" type, and
      • multiple IP address pools in the same IP family, e.g., IPv4 or IPv6 in the cluster's network.

      Show the network type: OVNKubernetes

      $ oc get network.config/cluster -o yaml | yq -r '.spec.networkType'
OVNKubernetes

      In this example, there are two IPv4 IP address pools:

      $ oc get network.config/cluster -o yaml | yq -y '.spec.clusterNetwork' 
- cidr: 10.128.0.0/14
  hostPrefix: 23
- cidr: 10.254.0.0/16
  hostPrefix: 24

      A cluster with only one cluster network configured is not affected.
      A cluster with standard dual-stack network is not affected.

      What is the impact? Is it serious enough to warrant removing update recommendations?

      • Upgrade will fail with the ovnkube-node pods stuck in CrashLoopBackOff state. Those pods can be found with the following command: 
        $ oc get pod -n openshift-ovn-kubernetes -l app=ovnkube-node
      • The cluster is in a bad state where new pods cannot be created successfully.

      How involved is remediation?

      • There is no remediation yet.

      Is this a regression?

      • Yes. Upgrade to 4.18.z

       

            [CORENET-5832] Impact statement request for OCPBUGS-49994 ovnkube-controller stuck in crashloopbackoff when upgrading to 4.18 with clusterNetwork set to multiple networks of the same IP family

            W. Trevor King added a comment -

            ...can we remove this for 4.18.9 and above?

            Already done last week via graph-data#7067.

            W. Trevor King added a comment - ...can we remove this for 4.18.9 and above? Already done last week via graph-data#7067 .

            Martin Kennelly added a comment -

            hongkliu trking  can we remove this for 4.18.9 and above? https://amd64.ocp.releases.ci.openshift.org/releasestream/4-stable/release/4.18.9

             

            See "OCPBUGS-54671: EIP no reroute: allow multiple entries for pod to pod and pod to join #2510"

            Martin Kennelly added a comment - hongkliu trking   can we remove this for 4.18.9 and above? https://amd64.ocp.releases.ci.openshift.org/releasestream/4-stable/release/4.18.9   See " OCPBUGS-54671 : EIP no reroute: allow multiple entries for pod to pod and pod to join #2510"

            Hongkai Liu added a comment -

            Based on the impact assessment, known issue / conditional risk for this bug was added to the update graph. UpgradeBlocker label was added to this card for searchability.

            This card has been closed. Note this does not mean the bug is resolved, only that its impact is understood enough for setting up a conditional risk in the update graph. Please refer to OCPBUGS-49994 and its clones for information about fix state in particular versions.

            Details of the conditional risk:

            • Name: OVNWithMultipleClusterNetworks
            • Summary: The cluster upgrade cannot be complete and new pods cannot be created successfully if the cluster network configuration is with:
            • "OVNKubernetes" type, and
            • multiple IP address pools in the same IP family, e.g., IPv4 or IPv6 in the cluster's network.

            Hongkai Liu added a comment - Based on the impact assessment, known issue / conditional risk for this bug was added to the update graph. UpgradeBlocker label was added to this card for searchability. This card has been closed. Note this does not mean the bug is resolved, only that its impact is understood enough for setting up a conditional risk in the update graph. Please refer to OCPBUGS-49994 and its clones for information about fix state in particular versions. Details of the conditional risk: Name: OVNWithMultipleClusterNetworks Summary: The cluster upgrade cannot be complete and new pods cannot be created successfully if the cluster network configuration is with: "OVNKubernetes" type, and multiple IP address pools in the same IP family, e.g., IPv4 or IPv6 in the cluster's network.

              jcaamano@redhat.com Jaime Caamaño Ruiz
              hongkliu Hongkai Liu
              Votes:
              0 Vote for this issue
              Watchers:
              8 Start watching this issue

                Created:
                Updated:
                Resolved: