[
https://issues.apache.org/jira/browse/FLINK-5861?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15894285#comment-15894285
]
ASF GitHub Bot commented on FLINK-5861:
---------------------------------------
Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/3395#discussion_r104135197
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/taskexecutor/TaskExecutor.java
---
@@ -1114,6 +1122,12 @@ public void run() {
public void updateTaskExecutionState(final TaskExecutionState
taskExecutionState) {
TaskExecutor.this.updateTaskExecutionState(jobMasterLeaderId, jobMasterGateway,
taskExecutionState);
}
+
+ @Override
+ public void notifyJobManagerConnectionChanged(JobMasterGateway
jobMasterGateway, UUID jobMasterLeaderID) {
--- End diff --
This call can happen concurrently. You would have to synchronize the field
writes.
> TaskManager's components support updating JobManagerConnection
> --------------------------------------------------------------
>
> Key: FLINK-5861
> URL: https://issues.apache.org/jira/browse/FLINK-5861
> Project: Flink
> Issue Type: Sub-task
> Components: Distributed Coordination, JobManager
> Reporter: Biao Liu
> Assignee: Biao Liu
> Fix For: 1.3.0
>
>
> Some components in TaskManager, such as TaskManagerActions,
> CheckpointResponder, ResultPartitionConsumableNotifier,
> PartitionProducerStateChecker, need to support updating JobManagerConnection.
> So when JobManager fails and recovers, the tasks who keep old
> JobManagerConnection can be notified to update JobManagerConnection. The
> tasks can continue doing their jobs without failure.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
