You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Lei Yang (Jira)" <ji...@apache.org> on 2024/10/21 07:44:00 UTC

[jira] [Created] (FLINK-36575) ExecutionVertexInputInfo supports consuming subpartition groups

Lei Yang created FLINK-36575:
--------------------------------

             Summary: ExecutionVertexInputInfo supports consuming subpartition groups
                 Key: FLINK-36575
                 URL: https://issues.apache.org/jira/browse/FLINK-36575
             Project: Flink
          Issue Type: Sub-task
          Components: Runtime / Coordination
            Reporter: Lei Yang


Currently, the ExecutionVertexInputInfo describes the task's input through a combination of a PartitionRange and a SubpartitionRange. However, in the case of skewed join optimization, we need to split a group of data corresponding to the specific key, which may result in a downstream task subscribing to multiple combinations of PartitionRanges and SubpartitionRanges. 

Therefore, we need to modify the ExecutionVertexInputInfo to describe the input data as multiple combinations of PartitionRanges and SubpartitionRanges to meet the requirements of the aforementioned scenario and improve Flink's flexibility in describing the task's inputs.



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