azagrebin commented on a change in pull request #12256: URL: https://github.com/apache/flink/pull/12256#discussion_r437991598
########## File path: flink-runtime/src/main/java/org/apache/flink/runtime/scheduler/OneSlotPerExecutionSlotAllocator.java ########## @@ -0,0 +1,277 @@ +/* + * 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; + +import org.apache.flink.api.common.time.Time; +import org.apache.flink.runtime.clusterframework.types.AllocationID; +import org.apache.flink.runtime.clusterframework.types.SlotProfile; +import org.apache.flink.runtime.concurrent.FutureUtils; +import org.apache.flink.runtime.jobmanager.scheduler.Locality; +import org.apache.flink.runtime.jobmaster.LogicalSlot; +import org.apache.flink.runtime.jobmaster.SlotOwner; +import org.apache.flink.runtime.jobmaster.SlotRequestId; +import org.apache.flink.runtime.jobmaster.slotpool.PhysicalSlot; +import org.apache.flink.runtime.jobmaster.slotpool.PhysicalSlotRequest; +import org.apache.flink.runtime.jobmaster.slotpool.SingleLogicalSlot; +import org.apache.flink.runtime.jobmaster.slotpool.SlotProvider; +import org.apache.flink.runtime.scheduler.strategy.ExecutionVertexID; +import org.apache.flink.runtime.taskmanager.TaskManagerLocation; +import org.apache.flink.util.FlinkException; + +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.util.ArrayList; +import java.util.Collection; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.concurrent.CompletableFuture; +import java.util.function.Function; +import java.util.stream.Collectors; + +import static org.apache.flink.util.Preconditions.checkNotNull; +import static org.apache.flink.util.Preconditions.checkState; + +/** + * This slot allocator will request one physical slot for each single execution vertex. + * The slots will be requested in bulks so that the {@link SlotProvider} can check + * whether this bulk of slot requests can be fulfilled at the same time. + * It has several limitations: + * + * <p>1. Slot sharing will be ignored. + * + * <p>2. Co-location constraints are not allowed. + * + * <p>3. Intra-bulk input location preferences will be ignored. + */ +public class OneSlotPerExecutionSlotAllocator extends AbstractExecutionSlotAllocator { + + private static final Logger LOG = LoggerFactory.getLogger(OneSlotPerExecutionSlotAllocator.class); + + private final SlotProvider slotProvider; + + private final SlotOwner slotOwner; + + private final boolean slotWillBeOccupiedIndefinitely; + + private final Time allocationTimeout; + + public OneSlotPerExecutionSlotAllocator( + final SlotProvider slotProvider, + final InputsLocationsRetriever inputsLocationsRetriever, + final boolean slotWillBeOccupiedIndefinitely, + final Time allocationTimeout) { + + super(inputsLocationsRetriever); + + this.slotProvider = checkNotNull(slotProvider); + this.slotWillBeOccupiedIndefinitely = slotWillBeOccupiedIndefinitely; + this.allocationTimeout = checkNotNull(allocationTimeout); + + this.slotOwner = new OneSlotPerExecutionSlotAllocatorSlotOwner(); + } + + @Override + public List<SlotExecutionVertexAssignment> allocateSlotsFor( + final List<ExecutionVertexSchedulingRequirements> executionVertexSchedulingRequirements) { + + validateSchedulingRequirements(executionVertexSchedulingRequirements); + + validateNoCoLocationConstraint(executionVertexSchedulingRequirements); + + final Set<ExecutionVertexID> allExecutionVertexIds = executionVertexSchedulingRequirements.stream() + .map(ExecutionVertexSchedulingRequirements::getExecutionVertexId) + .collect(Collectors.toSet()); + + final Map<ExecutionVertexID, SlotRequestId> executionVertexSlotRequestIds = + generateExecutionVertexSlotRequestIds(allExecutionVertexIds); + + final List<SlotExecutionVertexAssignment> slotExecutionVertexAssignments = + createSlotExecutionVertexAssignments(executionVertexSchedulingRequirements, executionVertexSlotRequestIds); + + final List<CompletableFuture<PhysicalSlotRequest>> physicalSlotRequestFutures = + createPhysicalSlotRequestFutures( + executionVertexSchedulingRequirements, + allExecutionVertexIds, + executionVertexSlotRequestIds); + + allocateSlotsForAssignments( + physicalSlotRequestFutures, + slotExecutionVertexAssignments, + executionVertexSlotRequestIds); + + return slotExecutionVertexAssignments; Review comment: Are you worried about performance for `new ArrayList<>(slotExecutionVertexAssignments.values())` to make `slotExecutionVertexAssignments` a map? Then you would not need `allExecutionVertexIds` in `createPhysicalSlotRequestFutures` as it could be just `slotExecutionVertexAssignments.keySet()`. Also not sure, why the API is to return `List<SlotExecutionVertexAssignment>`. Do we need list indexing anywhere on the using side? Otherwise, `allocateSlotsFor` could return `Collection`, then it could return just `slotExecutionVertexAssignments.values()`. ---------------------------------------------------------------- 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: us...@infra.apache.org