Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-7359

[Azure] Replace master failed as new master did not add into lb backend

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Critical Critical
    • None
    • 4.13
    • Cloud Compute / Unknown
    • None
    • Critical
    • None
    • CLOUD Sprint 232, CLOUD Sprint 233, CLOUD Sprint 234
    • 3
    • Approved
    • False
    • Hide

      This is a regression in behaviour from 4.12

      Show
      This is a regression in behaviour from 4.12
    • Hide
      The Kubernetes 1.26 release introduced changes to the node infrastructure, such as removing an unhealthy node with an `NotReady` status from the public load balancer to prevent the node from receiving routing traffic. These changes can impact a {product-title} node that runs inside a cluster on Microsoft Azure, because removing a node from a public load balancer can result in the node losing internet connectivity. This issue might render the node unable to regain a ready status and establish an outbound connection. The {product-title} {product-version} release fixes this issue so that a node marked with `NotReady` status is detected by the `kube-proxy` based health probes, now the default for services of `type=LoadBalancer` with `externalTrafficPolicy=Cluster`, without the need of node detachment from the public load balancer. This means that a node can retain an outbound internet connection throughout these phases. (link:https://issues.redhat.com/browse/OCPBUGS-7359[*OCPBUGS-7359*])
      Show
      The Kubernetes 1.26 release introduced changes to the node infrastructure, such as removing an unhealthy node with an `NotReady` status from the public load balancer to prevent the node from receiving routing traffic. These changes can impact a {product-title} node that runs inside a cluster on Microsoft Azure, because removing a node from a public load balancer can result in the node losing internet connectivity. This issue might render the node unable to regain a ready status and establish an outbound connection. The {product-title} {product-version} release fixes this issue so that a node marked with `NotReady` status is detected by the `kube-proxy` based health probes, now the default for services of `type=LoadBalancer` with `externalTrafficPolicy=Cluster`, without the need of node detachment from the public load balancer. This means that a node can retain an outbound internet connection throughout these phases. (link: https://issues.redhat.com/browse/OCPBUGS-7359 [* OCPBUGS-7359 *])
    • Bug Fix
    • Done

      Description of problem:

      On azure, delete a master, old machine stuck in Deleting, some pods in cluster are in ImagePullBackOff, check from azure console, new master did not add into lb backend, seems this lead the machine has no internet connection.

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

      4.13.0-0.nightly-2023-02-12-024338

      How reproducible:

      Always

      Steps to Reproduce:

      1. Set up a cluster on Azure, networkType ovn
2. Delete a master
3. Check master and pod

      Actual results:

      Old machine stuck in Deleting,  some pods are in ImagePullBackOff.
 $ oc get machine    
NAME                                    PHASE      TYPE              REGION   ZONE   AGE
zhsunaz2132-5ctmh-master-0              Deleting   Standard_D8s_v3   westus          160m
zhsunaz2132-5ctmh-master-1              Running    Standard_D8s_v3   westus          160m
zhsunaz2132-5ctmh-master-2              Running    Standard_D8s_v3   westus          160m
zhsunaz2132-5ctmh-master-flqqr-0        Running    Standard_D8s_v3   westus          105m
zhsunaz2132-5ctmh-worker-westus-dhwfz   Running    Standard_D4s_v3   westus          152m
zhsunaz2132-5ctmh-worker-westus-dw895   Running    Standard_D4s_v3   westus          152m
zhsunaz2132-5ctmh-worker-westus-xlsgm   Running    Standard_D4s_v3   westus          152m

$ oc describe machine zhsunaz2132-5ctmh-master-flqqr-0  -n openshift-machine-api |grep -i "Load Balancer"
      Internal Load Balancer:  zhsunaz2132-5ctmh-internal
      Public Load Balancer:      zhsunaz2132-5ctmh

$ oc get node            
NAME                                    STATUS     ROLES                  AGE    VERSION
zhsunaz2132-5ctmh-master-0              Ready      control-plane,master   165m   v1.26.0+149fe52
zhsunaz2132-5ctmh-master-1              Ready      control-plane,master   165m   v1.26.0+149fe52
zhsunaz2132-5ctmh-master-2              Ready      control-plane,master   165m   v1.26.0+149fe52
zhsunaz2132-5ctmh-master-flqqr-0        NotReady   control-plane,master   109m   v1.26.0+149fe52
zhsunaz2132-5ctmh-worker-westus-dhwfz   Ready      worker                 152m   v1.26.0+149fe52
zhsunaz2132-5ctmh-worker-westus-dw895   Ready      worker                 152m   v1.26.0+149fe52
zhsunaz2132-5ctmh-worker-westus-xlsgm   Ready      worker                 152m   v1.26.0+149fe52
$ oc describe node zhsunaz2132-5ctmh-master-flqqr-0
  Warning  ErrorReconcilingNode       3m5s (x181 over 108m)  controlplane         [k8s.ovn.org/node-chassis-id annotation not found for node zhsunaz2132-5ctmh-master-flqqr-0, macAddress annotation not found for node "zhsunaz2132-5ctmh-master-flqqr-0" , k8s.ovn.org/l3-gateway-config annotation not found for node "zhsunaz2132-5ctmh-master-flqqr-0"]

$ oc get po --all-namespaces | grep ImagePullBackOf   
openshift-cluster-csi-drivers                      azure-disk-csi-driver-node-l8ng4                                  0/3     Init:ImagePullBackOff   0              113m
openshift-cluster-csi-drivers                      azure-file-csi-driver-node-99k82                                  0/3     Init:ImagePullBackOff   0              113m
openshift-cluster-node-tuning-operator             tuned-bvvh7                                                       0/1     ImagePullBackOff        0              113m
openshift-dns                                      node-resolver-2p4zq                                               0/1     ImagePullBackOff        0              113m
openshift-image-registry                           node-ca-vxv87                                                     0/1     ImagePullBackOff        0              113m
openshift-machine-config-operator                  machine-config-daemon-crt5w                                       1/2     ImagePullBackOff        0              113m
openshift-monitoring                               node-exporter-mmjsm                                               0/2     Init:ImagePullBackOff   0              113m
openshift-multus                                   multus-4cg87                                                      0/1     ImagePullBackOff        0              113m
openshift-multus                                   multus-additional-cni-plugins-mc6vx                               0/1     Init:ImagePullBackOff   0              113m
openshift-ovn-kubernetes                           ovnkube-master-qjjsv                                              0/6     ImagePullBackOff        0              113m
openshift-ovn-kubernetes                           ovnkube-node-k8w6j                                                0/6     ImagePullBackOff        0              113m

      Expected results:

      Replace master successful

      Additional info:

      Tested payload 4.13.0-0.nightly-2023-02-03-145213, same result.
Before we have tested in 4.13.0-0.nightly-2023-01-27-165107, all works well.

        1. backend pools.png
          backend pools.png
          249 kB

              ddonati@redhat.com Damiano Donati
              rhn-support-zhsun Zhaohua Sun
              Zhaohua Sun Zhaohua Sun
              Darragh Fitzmaurice Darragh Fitzmaurice
              Riccardo Ravaioli
              Votes:
              0 Vote for this issue
              Watchers:
              12 Start watching this issue

                Created:
                Updated:
                Resolved: