Powering On a cluster

Last Updated : Mar 20, 2026 |
Prolog information
Use this procedure to gracefully power on your solution cluster. You can also use this procedure to power on after an unexpected power down. Complete this procedure during a maintenance window.
Note:
Skip this procedure if you are upgrading from 4.3.1.0
  1. Log in to vCenter as an administrator or with the account used to deploy the cluster.
  2. Click the VMs and Templates tab.
  3. Power on the disk bearing node that was previously running the image-registry pod from step 4 in the powering down procedure. If the registry was running on a diskless node, then power on disk bearing node 1. Wait for 5 minutes before proceeding.
  4. Power on the remaining disk bearing node. Wait 5 minutes before proceeding.
  5. Power on the diskless node last. Wait 5 minutes before proceeding.
  6. Reconnect the SDS satellite pods.
    1. Run kubectl exec --namespace=piraeus deployment/piraeus-op-piraeus-operator-cs-controller linstor node list.
      ADDITIONAL INFORMATION:
      Note:
      This lists the FQDNs of each node that are needed to complete the next step.
    2. Run kubectl exec --namespace=piraeus deployment/piraeus-op-piraeus-operator-cs-controller linstor node reconnect <fqdn of node 1>.
    3. Repeat the above step using the FQDN for node 2 and node 3. Wait 5 minutes before proceeding.
  7. Start scaling up resources by first bringing up the Authorization pods using post-infra-upgrade. Wait for 20 minutes before continuing to the next step.
  8. Proceed when pod state has stabilized (not all pods will be running at this point because they are dependent on the orca-ref-input-adaptor pods.
  9. Scale up orca resource orca-ref-input-adaptor. Wait for another 90 minutes for the remaining pods to come up.
    ADDITIONAL INFORMATION:
    Important:
    If pods are still not running, see the below note and do not delete pods.
    Note:
    Do not use kubectl delete pod command. If there is a PV volume attachment associated to the pod, the risk of PV mount failures when deleting a pod significantly increases. Instead of deleting pods, scale down resources by executing kubectl scale <resource> <name> -n <namespace> --replicas=0where resource=statefulset, deployment and name = exact resource to scale (Example mstr-srv). Wait for the pod(s) to delete and then run kubectl scale <resource> <name> -n <namespace> --replicas=<former # of replicas>. If for any reason the pod is not deleted, wait approximately for 30 minutes. If the pods are not yet deleted and do not disappear, then escalate to the next level service team. Additional steps are needed to clean up the volume attachments in the case a deleted pod does not fully exit.
    Note:
    Avaya Analytics™: If CSP cluster certificate rotation has to be executed:
    • Do not run ccm cluster-rotate-certificates outside of a screen session.
    • If the cluster certificate rotation fails for any reason, do not reboot the cluster.
  10. Run ccm smoke-test.
    ADDITIONAL INFORMATION:
    Resolve any negative condition before proceeding.
  11. Perform basic sanity of Avaya Analytics™ to ensure it is working fine.