Hello everyone, Do you have flink-jobs helmchart r...
# random
v
Hello everyone, Do you have flink-jobs helmchart repo which is compatible with this flink-kubernetes-operator or any example helmchart job deployment Please suggest it will be really helpful Thanks a lot
g
We don’t have helm charts for job examples, but you can find the example yamls here: https://github.com/apache/flink-kubernetes-operator/tree/main/examples
v
Thanks
m
@vasanth narala right now on our team we're doing simple "template" flink deployment style helm charts simple ex from a previous POC
Copy code
apiVersion: <http://flink.apache.org/v1beta1|flink.apache.org/v1beta1>
kind: FlinkDeployment
metadata:
  name: {{ .Values.metadata.name }}
spec:
  image: {{ .Values.flink.container.image }}
  flinkVersion: {{ .Values.flink.version }}
  flinkConfiguration:
    taskmanager.numberOfTaskSlots: "{{ .Values.taskManager.numberOfTaskSlots }}"
  {{- if .Values.serviceAccount.create }}
  serviceAccount: {{ .Values.serviceAccount.name }}
  {{ end -}}
  jobManager:
    resource:
      memory: {{ .Values.jobManager.resource.memory }}
      cpu: {{ .Values.jobManager.resource.cpu }}
  taskManager:
    resource:
      memory: {{ .Values.taskManager.resource.memory }}
      cpu: {{ .Values.taskManager.resource.cpu }}
  job:
    jarURI: {{ .Values.job.jarURI }}
    parallelism: {{ .Values.job.parallelism }}
    upgradeMode: {{ .Values.job.upgradeMode }}
It doesn't encompass all the scenarios (like the job uri isn't optional so this wouldn't work for a session cluster)
Copy code
metadata:
  name: "basic-flink-example"
flink:
  version: "v1_15"
  container:
    image: "flink:1.15"
jobManager:
  resource:
    memory: "2048m"
    cpu: 1
taskManager:
  numberOfTaskSlots: "2"
  resource:
    memory: "2048m"
    cpu: 1
job:
  jarURI: local:///opt/flink/examples/streaming/StateMachineExample.jar
  parallelism: 2
  upgradeMode: stateless

serviceAccount:
  create: true
  name: flink-basic-example-sa
  namespace: default
v
Thanks @Marquis C If you have Github repo can you please provide will track the changes
g
The value here I feel is very limited because you are making almost every spec field configurable essentially moving the entire FlinkDeployment schema to the values .yaml
It's basically the same as writing the entire FlinkDeployment yaml with placeholders but without the possibility of making changes
v
ok
m
I think it’s a net benefit to leverage helm to help with the “package management” aspects, but agreed the simple example isn’t really super valuable.
Internally we couple this with, creating service accounts with perms to s3 roles and other ancillary tidbits that don’t come directly from the crd.