[GitHub] [flink] GJL commented on a change in pull request #8309: [FLINK-12229] [runtime] Implement LazyFromSourcesScheduling Strategy

[GitBox <gi...@apache.org>]

GJL commented on a change in pull request #8309: [FLINK-12229] [runtime] Implement LazyFromSourcesScheduling Strategy
URL: https://github.com/apache/flink/pull/8309#discussion_r280703833
 
 

 ##########
 File path: flink-runtime/src/main/java/org/apache/flink/runtime/scheduler/strategy/LazyFromSourcesSchedulingStrategy.java
 ##########
 @@ -0,0 +1,227 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.runtime.scheduler.strategy;
+
+import org.apache.flink.api.common.InputDependencyConstraint;
+import org.apache.flink.runtime.execution.ExecutionState;
+import org.apache.flink.runtime.io.network.partition.ResultPartitionID;
+import org.apache.flink.runtime.jobgraph.IntermediateDataSet;
+import org.apache.flink.runtime.jobgraph.IntermediateDataSetID;
+import org.apache.flink.runtime.jobgraph.JobGraph;
+import org.apache.flink.runtime.scheduler.DeploymentOption;
+import org.apache.flink.runtime.scheduler.ExecutionVertexDeploymentOption;
+import org.apache.flink.runtime.scheduler.SchedulerOperations;
+
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.HashMap;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Map;
+import java.util.Optional;
+import java.util.Set;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/**
+ * {@link SchedulingStrategy} instance for batch job which schedule vertices when input data are ready.
+ */
+public class LazyFromSourcesSchedulingStrategy implements SchedulingStrategy {
+
+	private final SchedulerOperations schedulerOperations;
+
+	private final SchedulingTopology schedulingTopology;
+
+	private final DeploymentOption deploymentOption = new DeploymentOption(true);
+
+	private final JobGraph jobGraph;
+
+	private final Map<IntermediateDataSetID, SchedulingIntermediateDataSet> intermediateDataSets;
+
+	public LazyFromSourcesSchedulingStrategy(
+		SchedulerOperations schedulerOperations,
+		SchedulingTopology schedulingTopology,
+		JobGraph jobGraph) {
+		this.schedulerOperations = checkNotNull(schedulerOperations);
+		this.schedulingTopology = checkNotNull(schedulingTopology);
+		this.jobGraph = checkNotNull(jobGraph);
+		this.intermediateDataSets = new HashMap<>();
+	}
+
+	@Override
+	public void startScheduling() {
+		List<ExecutionVertexDeploymentOption> executionVertexDeploymentOptions = new LinkedList<>();
+		for (SchedulingVertex schedulingVertex : schedulingTopology.getVertices()) {
+			if (jobGraph.findVertexByID(schedulingVertex.getJobVertexId()).isInputVertex()) {
+				// schedule vertices without consumed result partition
+				executionVertexDeploymentOptions.add(
+					new ExecutionVertexDeploymentOption(schedulingVertex.getId(), deploymentOption));
+			}
+
+			Collection<SchedulingResultPartition> partitions = schedulingVertex.getProducedResultPartitions();
+			if (partitions != null) {
+				for (SchedulingResultPartition srp : partitions) {
+					SchedulingIntermediateDataSet sid = intermediateDataSets.computeIfAbsent(srp.getResultId(),
+						(key) -> new SchedulingIntermediateDataSet());
+					sid.increaseProducerCnt();
+				}
+			}
+		}
+		if (!executionVertexDeploymentOptions.isEmpty()) {
+			schedulerOperations.allocateSlotsAndDeploy(executionVertexDeploymentOptions);
+		}
+	}
+
+	@Override
+	public void restartTasks(Set<ExecutionVertexID> verticesNeedingRestart) {
+		List<ExecutionVertexDeploymentOption> executionVertexDeploymentOptions = new ArrayList<>(verticesNeedingRestart.size());
+		for (ExecutionVertexID executionVertexID : verticesNeedingRestart) {
+			executionVertexDeploymentOptions.add(
+				new ExecutionVertexDeploymentOption(executionVertexID, deploymentOption));
+		}
+		schedulerOperations.allocateSlotsAndDeploy(executionVertexDeploymentOptions);
+	}
+
+	@Override
+	public void onExecutionStateChange(ExecutionVertexID executionVertexId, ExecutionState executionState) {
+		Optional<SchedulingVertex> schedulingVertex = schedulingTopology.getVertex(executionVertexId);
+		if (!schedulingVertex.isPresent()) {
+			throw new IllegalStateException("can not find scheduling vertex for " + executionVertexId);
+		}
+
+		if (ExecutionState.RUNNING.equals(executionState)) {
+			List<ExecutionVertexDeploymentOption> executionVertexDeploymentOptions = new ArrayList<>();
+			for (SchedulingResultPartition srp : schedulingVertex.get().getProducedResultPartitions()) {
+				if (srp.getPartitionType().isPipelined()) {
+					for (SchedulingVertex sv : srp.getConsumers()) {
+						executionVertexDeploymentOptions.add(
+							new ExecutionVertexDeploymentOption(sv.getId(), deploymentOption));
+					}
+				}
+			}
+			schedulerOperations.allocateSlotsAndDeploy(executionVertexDeploymentOptions);
+		} else if (ExecutionState.FINISHED.equals(executionState)) {
+			List<SchedulingResultPartition> finishedPartitions = null;
+			for (SchedulingResultPartition srp : schedulingVertex.get().getProducedResultPartitions()) {
+				if (srp.getPartitionType().isBlocking()) {
+					SchedulingIntermediateDataSet intermediateDataSet = intermediateDataSets.get(srp.getResultId());
+					if (intermediateDataSet == null) {
+						throw new IllegalStateException("can not find scheduling intermediate dataset for " + srp.getResultId());
+					}
+					final int refCnt = intermediateDataSet.decrementNumberOfRunningProducersAndGetRemaining();
+					if (refCnt < 0) {
+						throw new IllegalStateException("Decremented number of unfinished producers below 0. "
+							+ "This is most likely a bug in the execution state/intermediate result "
+							+ "partition management.");
+					} else if (refCnt == 0) {
+						if (finishedPartitions == null) {
+							finishedPartitions = new LinkedList<>();
+						}
+						finishedPartitions.add(srp);
+					}
+				}
+			}
+
+			if (finishedPartitions != null) {
+				List<SchedulingVertex> toScheduleConsumers = new LinkedList<>();
+				for (SchedulingResultPartition finishedPartition : finishedPartitions) {
+					toScheduleConsumers.addAll(finishedPartition.getConsumers());
+				}
+				scheduleBlockingConsumers(toScheduleConsumers);
+			}
+		}
+	}
+
+	@Override
+	public void onPartitionConsumable(ExecutionVertexID executionVertexId, ResultPartitionID resultPartitionId) {
+		// No need to react to this notification.
+	}
+
+	private void scheduleBlockingConsumers(List<SchedulingVertex> vertices) {
+		List<ExecutionVertexDeploymentOption> executionVertexDeploymentOptions = new LinkedList<>();
+		for (SchedulingVertex schedulingVertex : vertices) {
+			final ExecutionState state = schedulingVertex.getState();
+
+			if (ExecutionState.CREATED.equals(state)) {
+				final InputDependencyConstraint inputDependency = jobGraph.findVertexByID(schedulingVertex
+					.getJobVertexId()).getInputDependencyConstraint();
+
+				boolean shouldSchedule = true;
+				if (!InputDependencyConstraint.ANY.equals(inputDependency)) {
+					for (SchedulingResultPartition srp : schedulingVertex.getConsumedResultPartitions()) {
+						SchedulingIntermediateDataSet sid = intermediateDataSets.get(srp.getResultId());
+						if (sid == null) {
+							throw new IllegalArgumentException("can not find schedule intermediate data set for " + srp);
+						}
+
+						if (!sid.areAllPartitionsFinished()) {
+							shouldSchedule = false;
+							break;
+						}
+					}
+				}
+
+				if (shouldSchedule) {
+					executionVertexDeploymentOptions.add(
+						new ExecutionVertexDeploymentOption(schedulingVertex.getId(), deploymentOption));
+				}
+			}
+		}
+
+		if (!executionVertexDeploymentOptions.isEmpty()) {
+			schedulerOperations.allocateSlotsAndDeploy(executionVertexDeploymentOptions);
+		}
+	}
+
+	/**
+	 * The factory for creating {@link LazyFromSourcesSchedulingStrategy}.
+	 */
+	public static class Factory implements SchedulingStrategyFactory {
+		@Override
+		public SchedulingStrategy getInstance(
+			SchedulerOperations schedulerOperations,
+			SchedulingTopology schedulingTopology,
+			JobGraph jobGraph) {
+			return new LazyFromSourcesSchedulingStrategy(schedulerOperations, schedulingTopology, jobGraph);
+		}
+	}
+
+	/**
+	 * Representation of {@link IntermediateDataSet}.
+	 */
+	class SchedulingIntermediateDataSet {
 
 Review comment:
   Should be `static` because `SchedulingIntermediateDataSet` does not access state from the scheduling strategy.

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services