JM & TM readiness probe

[Alexey Trenikhun]

Hello,

How can I define rediness fork8s job cluster deployments? I think for job manager, I can use REST API and check job status, but what about task manager? Is anyway to ask task manager Pod is it ready or not?

Thanks,
Alexey

[Andrey Zagrebin-5]

Hi Alexey,

As far as I know, TaskManager does not expose the REST API. ResourceManager redirects some REST calls to TaskManager [1]:

/taskmanagers/:taskmanagerid/metrics
/taskmanagers/:taskmanagerid/thread-dump

These calls may be not so lightweight. I do not know others or how you ask e.g. the state of TaskManager, it is all only over internal communication (currently akka).

Maybe Chesnay has a better idea.

We also use TCP port 6122 as TaskManager livenessProbe [2]. This might not qualify as a readiness probe though.

Best,

Andrey

[1] https://ci.apache.org/projects/flink/flink-docs-release-1.11/monitoring/rest_api.html#taskmanagers-taskmanagerid-metrics

[2] https://ci.apache.org/projects/flink/flink-docs-release-1.11/ops/deployment/kubernetes.html#session-cluster-resource-definitions

On Wed, Aug 12, 2020 at 4:50 AM Alexey Trenikhun <[hidden email]> wrote:

Hello,

How can I define rediness fork8s job cluster deployments? I think for job manager, I can use REST API and check job status, but what about task manager? Is anyway to ask task manager Pod is it ready or not?

Thanks,
Alexey