Uploaded image for project: 'Red Hat OpenStack Services on OpenShift'
  1. Red Hat OpenStack Services on OpenShift
  2. OSPRH-13951

octavia-operator unstable when using DCN network settings

XMLWordPrintable

    • 1
    • False
    • Hide

      None

      Show
      None
    • False
    • ?
    • octavia-operator-container-1.0.7-8
    • None
    • Hide
      .Fixed stability issue with Load-balancing service health manager in DCN mode
      Before this update, when you ran Load-balancing service (octavia) health manager pods in DCN mode, pods were randomly restarted by the operator.
      With this update, the random restarts do not occur.
      Show
      .Fixed stability issue with Load-balancing service health manager in DCN mode Before this update, when you ran Load-balancing service (octavia) health manager pods in DCN mode, pods were randomly restarted by the operator. With this update, the random restarts do not occur.
    • Bug Fix
    • Done
    • VANS-010
    • 1
    • Important

      To Reproduce Steps to reproduce the behavior:

      Deploy rhoso with octavia and enable multi-AZ management network:

      spec:
        octavia:
          template:
            lbMgmtNetwork:
              availabilityZoneCIDRs:
                az1: 172.34.0.0/16
                az2: 172.44.0.0/16
              createDefaultLbMgmtNetwork: false
      

      The CIDR of the AZs are passed to the octavia-healthmanager pods via env vars:

      $ oc get daemonsets.apps octavia-healthmanager  -o yaml | grep -A1 MGMT_CIDR
              - name: MGMT_CIDR
                value: 172.24.0.0/16
              - name: MGMT_CIDR0
                value: 172.34.0.0/16
              - name: MGMT_CIDR1
                value: 172.44.0.0/16

      The issue is that the order of those env vars may differ in each reconciliation loop, we may also get

      $ oc get daemonsets.apps octavia-healthmanager  -o yaml | grep -A1 MGMT_CIDR
               - name: MGMT_CIDR
                 value: 172.24.0.0/16
               - name: MGMT_CIDR0
                 value: 172.44.0.0/16
               - name: MGMT_CIDR1
                 value: 172.34.0.0/16

      when the order changes, that changes the input parameters of the daemonset and recreates the pods.

      This behavior is not 100% reproducible, and occurs randomly, it can trigger an infinite loop of pod recreation.

      Expected behavior

      • the input parameters should be stable and only update the daemonset when necessary

      Bug impact

      • octavia-healthmanager may be randomly restarted and make octavia unusable in DCN mode

      Known workaround

      • no

      Note

      • Octavia DCN is not officially supported in 18.0.4

              rhn-support-gthiemon Gregory Thiemonge
              rhn-support-gthiemon Gregory Thiemonge
              rhos-dfg-networking-squad-vans
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: