The Confluence Helm Chart documents:

confluence:
  # Port definitions
  #
  ports:

    # -- The port on which the Confluence container listens for Hazelcast traffic
    #
    hazelcast: 5701

However, changing the above value to a custom value such as Port 5801 only works for a single node. Scaling to additional nodes will result in nodes encountering a Cluster Panic and shutting down.

The above custom value updates Confluence to listen on Port 5801 but the discovery of other nodes still occurs on Port 5701 .

The cause of this is that Hazelcast Kubernetes Discovery API appears to utilise the HAZELCAST_KUBERNETES_SERVICE_PORT environment value for the Port number to discover other nodes. When this environment variable is not set, it defaults to 5701.

Suggestion:

Update the Helm Chart deployment to automatically set the HAZELCAST_KUBERNETES_SERVICE_PORT environment variable to the value in confluence.ports.hazelcast from the Helm chart values.

Workaround:

To update Confluence to a custom Hazelcast port value, also set the HAZELCAST_KUBERNETES_SERVICE_PORT environment variable in the Helm Chart values file. e.g.

confluence:
  ports:
    hazelcast: 5801

  additionalEnvironmentVariables:
    - name: HAZELCAST_KUBERNETES_SERVICE_PORT
      value: "5801"