[
https://issues.apache.org/jira/browse/FLINK-27748?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17599355#comment-17599355
]
Konstantin Knauf commented on FLINK-27748:
------------------------------------------
[~long jiang] Better late than never. You can also set the maxParallelism per
Operator and that will be respect by reative mode.
> AdaptiveScheduler should support operator fixed parallelism
> -----------------------------------------------------------
>
> Key: FLINK-27748
> URL: https://issues.apache.org/jira/browse/FLINK-27748
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Task
> Reporter: john
> Priority: Minor
>
> In the job topology, if the user specifies the concurrency of the operator,
> AdaptiveScheduler should support the operator's maximum parallelism equal to
> the user-specified parallelism during the scheduling process. And the minimum
> parallelism is equal to the number of slots available to the cluster.
> This is especially useful in certain scenarios,
> For example, the parallelism of an operator that consumes Kafka is specified
> to be equal to the number of partitions. Or you want to control the write
> rate of the operator, etc.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
