[ 
https://issues.apache.org/jira/browse/FLINK-30808?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17681514#comment-17681514
 ] 

Junrui Li edited comment on FLINK-30808 at 1/28/23 6:25 AM:
------------------------------------------------------------

The failure is because currently in StreamGraph#createActualEdge, if the 
partitioner is a ForwardPartitioner, then the operator parallelism of the 
upstream and downstream of the partitioner will be required to be consistent. 
However, for the ForwardForConsecutiveHashPartitioner(which is a subclass of 
ForwardPartitioner) used in this case, it is not required that the parallelism 
of upstream and downstream operators must be consistent. So we can fix this bug 
by fixing the StreamGraph#createActualEdge.


was (Author: JIRAUSER293716):
The failure is because currently in StreamGraph#createActualEdge, if the 
partitioner is a ForwardPartitioner, then the operator parallelism of the 
upstream and downstream of the partitioner will be required to be consistent. 
However, for the ForConsecutiveHashPartitioner used in this case, it is not 
required that the parallelism of upstream and downstream operators must be 
consistent. So we can fix this bug by fixing the StreamGraph#createActualEdge.

> MultipleInputITCase failed with AdaptiveBatch Scheduler
> -------------------------------------------------------
>
>                 Key: FLINK-30808
>                 URL: https://issues.apache.org/jira/browse/FLINK-30808
>             Project: Flink
>          Issue Type: Bug
>          Components: Runtime / Coordination
>    Affects Versions: 1.16.0, 1.17.0
>            Reporter: Junrui Li
>            Assignee: Junrui Li
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.17.0
>
>
> MultipleInputITCase#testRelatedInputs failed with AdaptiveBatch Scheduler.
> {code:java}
> java.lang.UnsupportedOperationException: Forward partitioning does not allow 
> change of parallelism. Upstream operation: Calc[10]-14 parallelism: 1, 
> downstream operation: HashJoin[15]-20 parallelism: 3 You must use another 
> partitioning strategy, such as broadcast, rebalance, shuffle or global. {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to