Instructions

• Make a Docker image with python3, java, unzip, and the Rama release archive (e.g. rama-1.1.0.zip) at /home/rama/. See helpers/Dockerfile for a minimal example meeting these requirements.
• Configure k8s to allow swap space. When updating modules, Rama launches a new worker process for the new module instance alongside each existing worker and transitions responsibilities between them. So there's a temporary need for additional memory during this period, and configuring swap space ensures the pod won't run out of memory. See https://kubernetes.io/blog/2023/08/24/swap-linux-beta/
• Zookeeper cluster must be launched separately (e.g. with bitnami zookeeper helm chart)
• zookeeper.servers must be filled in values.yaml with list of all Zookeeper server names
• Persistent volumes must be allocated for each Rama node. It's recommended to have one class for the conductor and another class for the supervisors. This will make it easier on future cluster upgrades when associating the volume with a new pod for the upgraded Rama version.