Configuring SQL job settings
If you need to further customize your SQL Stream job, you can add more advanced features to configure the job restarting method and time, threads for parallelism, sample behavior, exatly once processing and restoring from savepoint.
Before running a SQL query, you can configure advanced features by clicking on the Settings tabs at the SQL window.
- Execution Mode
- You can select the execution mode for the SQL job that can be a session or a per-job mode. For more information, see the Executing SQL jobs in production mode documentation.
- Job parallelism (threads)
- The number of threads to start to process the job. Each thread consumes a slot on the cluster. When the Job Parallelism is set to 1, the job consumes the least resources. If the data provired supports parallel reads, increasing the parallelism can raise the maximum throughput. For example, when using Kafka as a data provider, setting the parallelism to the equal number as the partitions of the topic can be a starting point for performance tuning.
- Restore From Savepoint
- You can enable or disable restoring a SQL job from a Flink savepoint after stopping it. The savepoint is saved under hdfs:///user/flink/savepoints by default.
- Sample Count
- The number of sample entries shown under the Results tab. To have an unlimited number of sample entries, add 0 to the Sample Count value.
- Sample Window Size
- The number of sample entries to keep in under the Results tab. To have an unlimited number of sample entries, add 0 to the Sample Window Size value.
- Sample Behavior
- You have the following options to choose the behavior of the sampled data under the
- Sample all messages
- Sample one message every second
- Sample one message every five seconds
- Enable Checkpointing
- You can enable and disable checkpointing for a SQL job. By default the checkpointing is enabled.
- Checkpoint Mode
- Switching between checkpointing modes. You can choose between At Least Once or Exactly Once.
- Checkpoint Interval
- The time in milliseconds between checkpointing attempts.
- Checkpoint Timeout
- The maximum time in milliseconds until a checkpointing attempt is timed out.
- Tolerable Checkpoint Failures
- The number of checkpointing attempts until the job is aborted.
- Failure Restart Strategy
- Switching between restarting strategies when checkpointing is failed. You can choose between enabling and disabling auto recovery. By default auto recovery is enabled for checkpointing.