apache-flinkflink-sqlflink-table-api

Flink failed to trigger checkpoint when using table API


My flink streaming application (v1.14.4) contain JDBC connector used for initial fetch data from MySQL server Logic:

  1. JDBC table source -> select.where() -> convert to datastream
  2. Kafka datastream join jdbc table -> further computation

When I run the application locally I can see following exception

14:52:00.401 [Source: TableSourceScan(table=[[default_catalog, default_database, asset, project=[id, status]]], fields=[id, status]) -> Calc(select=[CAST(_UTF-16LE'sergey-test-asset-id':VARCHAR(2147483647) CHARACTER SET "UTF-16LE") AS id, status], where=[(id = _UTF-16LE'sergey-test-asset-id')]) -> TableToDataSteam(type=ROW<`id` STRING, `status` STRING> NOT NULL, rowtime=false) -> Sink: Print to Std. Out (4/4)#0] INFO  o.a.flink.runtime.taskmanager.Task - Source: TableSourceScan(table=[[default_catalog, default_database, asset, project=[id, status]]], fields=[id, status]) -> Calc(select=[CAST(_UTF-16LE'sergey-test-asset-id':VARCHAR(2147483647) CHARACTER SET "UTF-16LE") AS id, status], where=[(id = _UTF-16LE'sergey-test-asset-id')]) -> TableToDataSteam(type=ROW<`id` STRING, `status` STRING> NOT NULL, rowtime=false) -> Sink: Print to Std. Out (4/4)#0 (e8870cf296ac770346384fe2529b325f) switched from RUNNING to FINISHED.
...

14:57:52.963 [Checkpoint Timer] INFO  o.a.f.r.c.CheckpointCoordinator - Failed to trigger checkpoint for job 8303c423dd7b9e3f303f0b299d7d37bb because Some tasks of the job have already finished and checkpointing with finished tasks is not enabled. Failure reason: Not all required tasks are currently running.

I do understand that after SQL select statement flink mark jdbc operator to FINISHED state, but I need to continue run the streaming application and have checkpoints during runtime

Do I need to use execution.checkpointing.checkpoints-after-tasks-finish.enabled: true property to fix my issue OR do I need to change my application graph?


Solution

  • Yes, you need to use

    execution.checkpointing.checkpoints-after-tasks-finish.enabled: true
    

    so that the job can checkpoint despite the fact that the jdbc source tasks have completed.