[
https://issues.apache.org/jira/browse/FLINK-26628?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17506393#comment-17506393
]
Igal Shilman commented on FLINK-26628:
--------------------------------------
[~Fil Karnicki], I'm not sure that it has anything to do with StateFun.
I think that this is the case with any DataStream application and the
difference between submitting via the cli vs the REST api.
Let's double check with other members of the community that might be more
knowledge here, for example: [[email protected]]
> Use REST program arguments in StatefulFunctionsJob
> --------------------------------------------------
>
> Key: FLINK-26628
> URL: https://issues.apache.org/jira/browse/FLINK-26628
> Project: Flink
> Issue Type: Improvement
> Components: Stateful Functions
> Reporter: Fil Karnicki
> Priority: Major
>
> Currently the Program Arguments passed into the REST api don't get used by
> StreamingExecutionEnvironment in StatefulFunctionsJob (in this case the
> checkpointing will not be set)
> {code:java}
> --execution.checkpointing.interval 1000 --state.backend rocksdb
> --state.checkpoint-storage filesystem --state.checkpoints.dir file:///tmp/
> --statefun.embedded true {code}
> Conversely, Flink CLI params *do* get used by the
> StreamingExecutionEnvironment in statefun jobs
> {code:java}
> flink run -Dexecution.checkpointing.interval=1000 -Dstate.backend=rocksdb
> -Dstate.checkpoint-storage=filesystem -Dstate.checkpoints.dir=file:///tmp/
> -Dstatefun.embedded=true myjar.jar{code}
>
> To reproduce,
> # clone and run mvn package on
> [https://github.com/FilKarnicki/statefun-flinkjob/tree/argsNotUsedViaRest]
> # run docker-compose up in flinkjob/docker-compose
> # observe checkpointing happening for this job
> # go to [http://localhost:8081/#/submit] and submit
> flinkjob-1.0-SNAPSHOT.jar again manually from target with program arguments
> {code:java}
> --execution.checkpointing.interval 1000 --state.backend rocksdb
> --state.checkpoint-storage filesystem --state.checkpoints.dir file:///tmp/
> --statefun.embedded true {code}
> 5. observe no checkpointing happening for the second job
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
