Apache Flink
CertifiedTasks that submit, monitor, cancel, and trigger savepoints for Apache Flink jobs over the REST API.
Provide the Flink JobManager REST restUrl plus job details such as jarUri, entryClass, args, parallelism, and optional savepoint paths to upload and run jobs, monitor status, cancel executions, or trigger savepoints directly from Kestra.
Apache Flink
Tasks that submit, monitor, cancel, and trigger savepoints for Apache Flink jobs over the REST API.
Provide the Flink JobManager REST restUrl plus job details such as jarUri, entryClass, args, parallelism, and optional savepoint paths to upload and run jobs, monitor status, cancel executions, or trigger savepoints directly from Kestra.
tasks
How to use the Apache Flink plugin
Submit and manage Flink jobs via the JobManager REST API from Kestra flows.
Authentication
Set restUrl to the Flink JobManager REST endpoint (e.g. http://jobmanager: 8081) on each task, or apply it globally with plugin defaults. The plugin communicates over plain HTTP — no built-in auth fields. If your cluster requires authentication, place a reverse proxy in front of the JobManager.
Tasks
Submit uploads a JAR and submits a job — set restUrl, jarUri, and entryClass (all required). Optionally pass program args, set parallelism, provide extra Flink configuration via jobConfig, or restore from a savepoint with restoreFromSavepoint.
SubmitSql executes a SQL statement via the Flink SQL Gateway — set gatewayUrl and statement (both required). Optionally set sessionName to reuse a named session and configure catalog/database via sessionConfig. Control execution timeout with statementTimeout (default 300 seconds).
CancelJob stops a running job by jobId — set restUrl and jobId (both required). Set withSavepoint: true and savepointDir to create a final savepoint before stopping. Set drainJob: true to flush buffered data before cancellation.
TriggerSavepoint creates a savepoint for a running job — set restUrl and jobId (both required). Optionally set targetDirectory and cancelJob: true to stop the job after the savepoint completes.
MonitorJob is a polling trigger — set restUrl and jobId (both required). It fires when the job reaches a terminal state. Control polling with interval (default 10 seconds). Set failOnError: false to treat a FAILED state as non-fatal.
