zhuzhurk commented on a change in pull request #15310:
URL: https://github.com/apache/flink/pull/15310#discussion_r603174108
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/scheduler/strategy/PipelinedRegionSchedulingStrategy.java
##########
@@ -92,15 +92,82 @@ private void init() {
}
}
+ private void initCrossRegionConsumerPartitionGroups() {
+ Set<ConsumedPartitionGroup> visitedPartitionGroups =
+ Collections.newSetFromMap(new IdentityHashMap<>());
+
+ for (SchedulingPipelinedRegion pipelinedRegion :
+ schedulingTopology.getAllPipelinedRegions()) {
+ for (ConsumedPartitionGroup consumedPartitionGroup :
+ pipelinedRegion.getAllBlockingConsumedPartitionGroups()) {
+ if (!visitedPartitionGroups.contains(consumedPartitionGroup)) {
+ visitedPartitionGroups.add(consumedPartitionGroup);
+
+ SchedulingPipelinedRegion producerRegion = null;
+ for (IntermediateResultPartitionID partitionId :
consumedPartitionGroup) {
+ SchedulingPipelinedRegion region =
getProducerRegion(partitionId);
+ if (producerRegion == null) {
+ producerRegion = region;
+ } else if (producerRegion != region) {
+
crossRegionConsumerPartitionGroups.add(consumedPartitionGroup);
+ }
+ }
+ }
+ }
+ }
+ }
+
+ private SchedulingPipelinedRegion
getProducerRegion(IntermediateResultPartitionID partitionId) {
+ return schedulingTopology.getPipelinedRegionOfVertex(
+
schedulingTopology.getResultPartition(partitionId).getProducer().getId());
+ }
+
+ private void initPartitionGroupConsumerRegions() {
+ for (SchedulingPipelinedRegion region :
schedulingTopology.getAllPipelinedRegions()) {
+ for (ConsumedPartitionGroup consumedPartitionGroup :
+ region.getAllBlockingConsumedPartitionGroups()) {
+ if
(crossRegionConsumerPartitionGroups.contains(consumedPartitionGroup)
+ ||
isExternalConsumedPartitionGroup(consumedPartitionGroup, region)) {
+ partitionGroupConsumerRegions
+ .computeIfAbsent(consumedPartitionGroup, group ->
new HashSet<>())
+ .add(region);
+ }
+ }
+ }
+ }
+
+ private void initCorrelatedResultPartitionGroups() {
+ for (ConsumedPartitionGroup consumedPartitionGroup :
+ partitionGroupConsumerRegions.keySet()) {
+ for (IntermediateResultPartitionID partitionId :
consumedPartitionGroup) {
+ correlatedResultPartitionGroups
+ .computeIfAbsent(
+ partitionId.getIntermediateDataSetID(), id ->
new HashSet<>())
+ .add(consumedPartitionGroup);
+ }
+ }
+ }
+
@Override
public void startScheduling() {
final Set<SchedulingPipelinedRegion> sourceRegions =
IterableUtils.toStream(schedulingTopology.getAllPipelinedRegions())
- .filter(region ->
!region.getConsumedResults().iterator().hasNext())
+ .filter(this::noExternalBlockingConsumedPartitions)
.collect(Collectors.toSet());
maybeScheduleRegions(sourceRegions);
}
+ private boolean
noExternalBlockingConsumedPartitions(SchedulingPipelinedRegion region) {
+ for (ConsumedPartitionGroup consumedPartitionGroup :
+ region.getAllBlockingConsumedPartitionGroups()) {
+ if
(crossRegionConsumerPartitionGroups.contains(consumedPartitionGroup)
Review comment:
Thanks for reporting this problem. Go ahead to open a new JIRA for it.
--
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]
