OpenStack known issues

This section lists the OpenStack known issues with workarounds for the Mirantis OpenStack for Kubernetes release 21.2.

[13422] Redis pods remain in Pending state and cause update failure

Fixed in MOS 21.3

During the MOS cluster update to Cluster release 6.14.0, some Redis pods may remain in the Pending state and cause update failure.

Workaround:

  1. Scale the Redis deployment to 0 replicas:

    kubectl -n openstack-redis scale deployment rfs-openstack-redis --replicas=0

    Wait for the pods removal.

  2. Scale the Redis deployment back to 3 replicas:

    kubectl -n openstack-redis scale deployment rfs-openstack-redis --replicas=3

  3. Obtain the list of replicas:

    kubectl -n openstack-redis get replicaset

    Example of system response:

    NAME                                         DESIRED  CURRENT  READY  AGE
os-redis-operator-redisoperator-6bd8455f8c   1        1        1      26h
rfs-openstack-redis-568b8f6fcb               0        0        0      26h
rfs-openstack-redis-798655cf9b               3        3        3      24h

  4. Remove the ReplicaSet with 0 replicas:

    kubectl -n openstack-redis delete replicaset rfs-openstack-redis-568b8f6fcb

[12511] Kubernetes workers remain in Prepare state

Fixed in MOS 21.3

During the MOS cluster update to Cluster release 6.14.0, Kubernetes nodes may get stuck in the Prepare state. At the same time, the LCM Controller logs may contain the following errors:

evicting pod "horizon-57f7ccff74-d469c"
error when evicting pod "horizon-57f7ccff74-d469c" (will retry after
5s): Cannot evict pod as it would violate the pod's disruption budget.

The workaround is to decrease the Pod Disruption Budget (PDB) limit for Horizon by executing the following command on the managed cluster:

kubectl -n openstack patch pdb horizon -p='{"spec": {"minAvailable": 1}}'

[13273] Octavia amphora may get stuck after cluster update

Fixed in MOS 21.4

After the MOS cluster update, Octavia amphora may get stuck with the WARNING octavia.amphorae.drivers.haproxy.rest_api_driver [-] Could not connect to instance. Retrying. error message present in the Octavia worker logs. The workaround is to manually switch the Octavia amphorae driver from V2 to V1.

Workaround:

  1. In the OsDpl CR, specify the following configuration:

    spec:
  services:
    load-balancer:
      octavia:
        values:
          conf:
            octavia:
              api_settings:
                default_provider_driver: amphora

  2. Trigger the OpenStack deployment to restart Octavia:

    kubectl apply -f openstackdeployment.yaml

    To monitor the status:

    kubectl -n openstack get pods
kubectl -n openstack describe osdpl osh-dev

[6912] Octavia load balancers may not work properly with DVR

Limitation

When Neutron is deployed in the DVR mode, Octavia load balancers may not work correctly. The symptoms include both failure to properly balance traffic and failure to perform an amphora failover. For details, see DVR incompatibility with ARP announcements and VRRP.