Below is each yaml document with are the yaml documents which describe the bare minimum infrastructure needed for a functional Flow server. It is best to start with a single-node proof of concept deployment. Once that works, the deployment can be extended to multi-node with elastic worker allocation. Each section is explained below.

...

apiVersion: v1
kind: Pod
metadata:
  name: flowheadnode
  namespace: partek-flow
  labels:
    app.kubernetes.io/name: flowheadnode
    deployment: dev
spec:
  securityContext:
    fsGroup: 1000
  containers:
    - name: flowheadnode
      image: xxxxxxxxxxxx.dkr.ecr.us-west-2.amazonaws.com/partek-flow:current-23.0809.22
      resources:
        requests:
          memory: "16Gi"
          cpu: 8
      env:
        - name: PARTEKLM_LICENSE_FILE
          value: "@flexlmserver"
        - name: PARTEK_COMMON_NO_TOTAL_LIMITS
          value: "1"
        - name: CATALINA_OPTS
          value: "-DFLOW_WORKER_MEMORY_MB=1024 -DFLOW_WORKER_CORES=2 -Djavax.net.ssl.trustStore=/etc/flowconfig/cacerts -Xmx14g"
      volumeMounts:
        - name: home-flow
          mountPath: /home/flow
        - name: flowconfig
          readOnly: true
          mountPath: "/etc/flowconfig"
  volumes:
    - name: home-flow
      persistentVolumeClaim:
        claimName: partek-flow-pvc
    - name: flowconfig
      secret:
        secretName: flowconfig

Pod metadata

On a Kubernetes kubernetes cluster, all Flow deployments are placed in their own namespace, for example namespace: partek-flow. The label app.kubernetes.io/name: flowheadnode allows binding of a service or used to target other kubernetes infrastructure to this headnode pod. The label deployment: dev allows us to run running multiple Flow instances in this namespace (dev, tst, uat, prd, etc) if needed and allows workers to connect to the correct headnode. For stronger isolation, running each Flow instance in its own namespace is optimal.  

...

The Flow docker image requires 1) a writable volume mounted to /home/flow 2) This volume needs to be readable and writable by UID:GID 1000:1000 3) For a multi-node setup, this volume needs to be cross mounted to all worker pods. In this case, the persistent volume would be backed by some network storage device such as EFS, NFS, or a mounted FileGateway. 

...

Partek Flow uses FlexLM for licensing. Currently we do not offer or have implemented any alternative. Values for this envar environment variable can be:

@flexlmserveraddress - an external flexlm server. We provide a Partek specific container image and detail a kubernetes deployment for this below. This license server can also live outside the kubernetes cluster - the only requirement being that it is network accessible. 
/home/flow/.partekflow/license/Partek.lic - Use this path exactly. This path is internal to the headnode container and is persisted on a mounted PVC. 

...

-Xmx14g - JVM heap size. If the internal worker is not used, set this to be a little less than the kubernetes memory resource request. If the internal worker is an use, and the intent is to stay with a single-node deployment, then set this to be ~ 25% of the kubernetes memory resource request, but no less than ~ 4GiB4 GiB. 

The Flow headnode service definition

...

The flowheadnode service is needed 1) So so that workers have a dns DNS name (flowheadnode) to connect to when they start and 2) so that we can attach an ingress route to make the Flow web interface accessible to end users. The app.kubernetes.io/name: flowheadnode selector is what binds this to the flowheadnode pod. 

...

The license from Partek must contain VENDOR parteklm PORT=27001 so the vendor port remains at 27001 in order to match the service definition above. Without this, this port is randomly set by flexlmFlexLM. 

This image is currently available from public.ecr.aws/partek-flow/kube-flexlm-server but this may change in the future. 

...