You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Andrey Zagrebin (JIRA)" <ji...@apache.org> on 2019/04/18 13:27:00 UTC

[jira] [Updated] (FLINK-12203) Refactor ResultPartitionManager to break tie with Task

     [ https://issues.apache.org/jira/browse/FLINK-12203?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Andrey Zagrebin updated FLINK-12203:
------------------------------------
    Description: At the moment, we have ResultPartitionManager.releasePartitionsProducedBy which uses indexing by task in network environment. These methods are eventually used only by Task which already knows its partitions so Task can use ResultPartition.fail(cause) and TaskExecutor.failPartition could directly use NetworkEnviroment.releasePartitions(Collection<ResultPartitionID>). This also requires that JM Execution sends produced partition ids instead of just ExecutionAttemptID.  (was: At the moment, we have ResultPartitionManager.releasePartitionsProducedBy which uses indexing by task in network environment. These methods are eventually used only by Task which already knows its partitions so Task and TE.failPartition could directly use NetworkEnviroment.releasePartition(ResultPartitionID). This also requires that JM Execution sends produced partition ids instead of just ExecutionAttemptID.)

> Refactor ResultPartitionManager to break tie with Task
> ------------------------------------------------------
>
>                 Key: FLINK-12203
>                 URL: https://issues.apache.org/jira/browse/FLINK-12203
>             Project: Flink
>          Issue Type: Sub-task
>            Reporter: Andrey Zagrebin
>            Assignee: Andrey Zagrebin
>            Priority: Major
>
> At the moment, we have ResultPartitionManager.releasePartitionsProducedBy which uses indexing by task in network environment. These methods are eventually used only by Task which already knows its partitions so Task can use ResultPartition.fail(cause) and TaskExecutor.failPartition could directly use NetworkEnviroment.releasePartitions(Collection<ResultPartitionID>). This also requires that JM Execution sends produced partition ids instead of just ExecutionAttemptID.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)