Run already deployed job on Flink Cluster using RestClusterClient

Question

I am trying to run already deployed job on Flink Cluster using Rest request.

I had success using a simple rest client

POST http://localhost:8081/v1/jars/13775a71-0723-4c62-979b-7e9a9de3a0dc_some.jar/run { "programArgsList" : ["test1", "test2"] }

But I would like to do it from Java and since there already is a RestClusterClient I would love to use that but it is poorly documented

https://ci.apache.org/projects/flink/flink-docs-master/api/java/org/apache/flink/client/program/rest/RestClusterClient.html

From what I see from the code it looks like there is no such possibility with this RestClusterClient

It looks like it is following closely features implemnted in CLI and documented here https://ci.apache.org/projects/flink/flink-docs-stable/ops/cli.html so any run command requires passing a jar with the job.

There is no possibility to run already deployed job with this client. Am I right?

Answer 1

Flink's cluster REST API has been designed to work with any REST client. The RestClusterClient which comes with the flink-clients module is intended for internal usage. However, you can also use it directly by instantiating it with a correctly set up Flink Configuration .

The Configuration is used to retrieve the cluster REST endpoint. Hence, if you configure the wrong REST address or the wrong HA mode, then the RestClusterClient might not be able to talk to the cluster. If you know where the flink-conf.yaml with which you have started the cluster is, then I would recommend to use GlobalConfiguration#loadConfiguration(configurationDirectory) to load it. This should give you a good starting point.

Once you have started the RestClusterClient you can interact with the cluster via

• submitJob : Submit a new job (only supported if you have a session cluster deployed)
• requestJobResult : Result of a job execution; if it's a streaming job, then this might never complete
• cancel : Cancel the given job
• triggerSavepoint : Trigger a savepoint and return its path
• listJobs : List all currently running jobs on the cluster
• and some more calls