When running as a container in Kubernetes, the tentacle container cannot be configured in listening mode

[coryreid]

Tentacle offers four options for hostname configuration during registration:

• PublicIP
• FQDN
• ComputerName
• Custom (Paired with CustomPublicHostName)

However, when standing up a listening worker in AKS, these all fail. PublicIP resolves to the Azure IP address (52.x.x.x), which isn't registerable within the cluster. The FQDN/ComputerName options utilize the hostname, which maps to the container name in the pod. However, these aren't routable through https:// from the Octopus server. Finally, Custom would work if it were possible to know the cluster IP at registration time, which isn't currently possible.

Running hostname -i on the container provides the internal cluster IP, which can be used to successfully register and route tasks to a listening tentacle. Being able to choose that internal IP would make listening tentacles function in Kubernetes in a similar fashion to other Octopus server registration workflows.

[lpodolak]

I know it's been 2 years after the original post, but I managed to configure tentacle in listening mode in K8S. The key is to use the "Custom" PublicHostNameConfiguration and specify pod's IP as the value for CustomPublicHostName, like below

- name: PublicHostNameConfiguration
  value: Custom
- name: CustomPublicHostName
  valueFrom:
    fieldRef:
      apiVersion: v1
      fieldPath: status.podIP

Another useful tip is to have tentacle automatically unregistering from Octopus Server once the pod is killed:

lifecycle:
  preStop: # Deregister tentacle from the server when pod is killed (e.g. during deployments)
    exec:
      command: ["sh", "-c", "tentacle deregister-worker --server=$ServerUrl --apiKey=$ServerApiKey"]