akalash commented on a change in pull request #15221:
URL: https://github.com/apache/flink/pull/15221#discussion_r603903691
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/Execution.java
##########
@@ -719,7 +721,9 @@ private void updatePartitionConsumers(final
IntermediateResultPartition partitio
// Consumer is deploying => cache the partition info which would be
// sent after switching to running
// ----------------------------------------------------------------
- if (consumerState == DEPLOYING || consumerState == RUNNING) {
+ if (consumerState == DEPLOYING
Review comment:
It looks like it indeed can be removed from here. But in general,
calling sendUpdatePartitionInfoRpcCall in both RECOVERING and RUNNING states
are not a mistake. So I decided to left it here but maybe I am wrong. Maybe,
@zentol can share his opinion about that?
##########
File path:
flink-runtime/src/test/java/org/apache/flink/runtime/scheduler/SchedulerTestingUtils.java
##########
@@ -224,16 +224,21 @@ public static void canceledExecution(
public static void setExecutionToRunning(
DefaultScheduler scheduler, JobVertexID jvid, int subtask) {
final ExecutionAttemptID attemptID = getAttemptId(scheduler, jvid,
subtask);
+ scheduler.updateTaskExecutionState(
Review comment:
I made this method more general.
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/taskmanager/Task.java
##########
@@ -505,6 +505,7 @@ AbstractInvokable getInvokable() {
public boolean isBackPressured() {
if (invokable == null
|| consumableNotifyingPartitionWriters.length == 0
+ || executionState != ExecutionState.RECOVERING
Review comment:
Yep, you absolutely right. My mistake.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
