Examples with SplitPlacementResult com.facebook.presto.execution.scheduler.SplitPlacementResult used on opensource projects

Example 1 with SplitPlacementResult

use of com.facebook.presto.execution.scheduler.SplitPlacementResult in project presto by prestodb.

the class SimpleNodeSelector method computeAssignments.

@Override
public SplitPlacementResult computeAssignments(Set<Split> splits, List<RemoteTask> existingTasks) {
    Multimap<InternalNode, Split> assignment = HashMultimap.create();
    NodeMap nodeMap = this.nodeMap.get().get();
    NodeAssignmentStats assignmentStats = new NodeAssignmentStats(nodeTaskMap, nodeMap, existingTasks);
    List<InternalNode> eligibleNodes = getEligibleNodes(maxTasksPerStage, nodeMap, existingTasks);
    NodeSelection randomNodeSelection = new RandomNodeSelection(eligibleNodes, minCandidates);
    Set<InternalNode> blockedExactNodes = new HashSet<>();
    boolean splitWaitingForAnyNode = false;
    NodeProvider nodeProvider = nodeMap.getActiveNodeProvider(nodeSelectionHashStrategy);
    OptionalInt preferredNodeCount = OptionalInt.empty();
    for (Split split : splits) {
        List<InternalNode> candidateNodes;
        switch(split.getNodeSelectionStrategy()) {
            case HARD_AFFINITY:
                candidateNodes = selectExactNodes(nodeMap, split.getPreferredNodes(nodeProvider), includeCoordinator);
                preferredNodeCount = OptionalInt.of(candidateNodes.size());
                break;
            case SOFT_AFFINITY:
                // Using all nodes for soft affinity scheduling with modular hashing because otherwise temporarily down nodes would trigger too much rehashing
                if (nodeSelectionHashStrategy == MODULAR_HASHING) {
                    nodeProvider = new ModularHashingNodeProvider(nodeMap.getAllNodes());
                }
                candidateNodes = selectExactNodes(nodeMap, split.getPreferredNodes(nodeProvider), includeCoordinator);
                preferredNodeCount = OptionalInt.of(candidateNodes.size());
                candidateNodes = ImmutableList.<InternalNode>builder().addAll(candidateNodes).addAll(randomNodeSelection.pickNodes(split)).build();
                break;
            case NO_PREFERENCE:
                candidateNodes = randomNodeSelection.pickNodes(split);
                break;
            default:
                throw new PrestoException(NODE_SELECTION_NOT_SUPPORTED, format("Unsupported node selection strategy %s", split.getNodeSelectionStrategy()));
        }
        if (candidateNodes.isEmpty()) {
            log.debug("No nodes available to schedule %s. Available nodes %s", split, nodeMap.getActiveNodes());
            throw new PrestoException(NO_NODES_AVAILABLE, "No nodes available to run query");
        }
        SplitWeight splitWeight = split.getSplitWeight();
        Optional<InternalNodeInfo> chosenNodeInfo = chooseLeastBusyNode(splitWeight, candidateNodes, assignmentStats::getTotalSplitsWeight, preferredNodeCount, maxSplitsWeightPerNode, assignmentStats);
        if (!chosenNodeInfo.isPresent()) {
            chosenNodeInfo = chooseLeastBusyNode(splitWeight, candidateNodes, assignmentStats::getQueuedSplitsWeightForStage, preferredNodeCount, maxPendingSplitsWeightPerTask, assignmentStats);
        }
        if (chosenNodeInfo.isPresent()) {
            split = new Split(split.getConnectorId(), split.getTransactionHandle(), split.getConnectorSplit(), split.getLifespan(), new SplitContext(chosenNodeInfo.get().isCacheable()));
            InternalNode chosenNode = chosenNodeInfo.get().getInternalNode();
            assignment.put(chosenNode, split);
            assignmentStats.addAssignedSplit(chosenNode, splitWeight);
        } else {
            if (split.getNodeSelectionStrategy() != HARD_AFFINITY) {
                splitWaitingForAnyNode = true;
            } else // Exact node set won't matter, if a split is waiting for any node
            if (!splitWaitingForAnyNode) {
                blockedExactNodes.addAll(candidateNodes);
            }
        }
    }
    ListenableFuture<?> blocked;
    if (splitWaitingForAnyNode) {
        blocked = toWhenHasSplitQueueSpaceFuture(existingTasks, calculateLowWatermark(maxPendingSplitsWeightPerTask));
    } else {
        blocked = toWhenHasSplitQueueSpaceFuture(blockedExactNodes, existingTasks, calculateLowWatermark(maxPendingSplitsWeightPerTask));
    }
    return new SplitPlacementResult(blocked, assignment);
}

Example 2 with SplitPlacementResult

use of com.facebook.presto.execution.scheduler.SplitPlacementResult in project presto by prestodb.

the class TestNodeScheduler method testAffinityAssignmentWithModularHashing.

@Test
public void testAffinityAssignmentWithModularHashing() {
    NodeTaskMap nodeTaskMap = new NodeTaskMap(finalizerService);
    TestingTransactionHandle transactionHandle = TestingTransactionHandle.create();
    NodeSchedulerConfig nodeSchedulerConfig = new NodeSchedulerConfig().setMaxSplitsPerNode(20).setIncludeCoordinator(false).setMaxPendingSplitsPerTask(10);
    NodeScheduler nodeScheduler = new NodeScheduler(new LegacyNetworkTopology(), nodeManager, new NodeSelectionStats(), nodeSchedulerConfig, nodeTaskMap, new ThrowingNodeTtlFetcherManager(), new NoOpQueryManager(), new SimpleTtlNodeSelectorConfig());
    NodeSelector nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 3);
    Set<Split> splits = new HashSet<>();
    // Adding one more split (1 % 3 = 1), 1 splits will be distributed to 1 nodes
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestAffinitySplitRemote(1)));
    SplitPlacementResult splitPlacementResult = nodeSelector.computeAssignments(splits, ImmutableList.of());
    Set<InternalNode> internalNodes = splitPlacementResult.getAssignments().keySet();
    assertEquals(internalNodes.size(), 1);
    // Adding one more split (2 % 3 = 2), 2 splits will be distributed to 2 nodes
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestAffinitySplitRemote(2)));
    splitPlacementResult = nodeSelector.computeAssignments(splits, getRemoteTableScanTask(splitPlacementResult));
    Set<InternalNode> internalNodesSecondCall = splitPlacementResult.getAssignments().keySet();
    assertEquals(internalNodesSecondCall.size(), 2);
    // adding one more split(4 % 3 = 1) that will fall into the same slots, 3 splits will be distributed to 2 nodes still
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestAffinitySplitRemote(4)));
    splitPlacementResult = nodeSelector.computeAssignments(splits, getRemoteTableScanTask(splitPlacementResult));
    assertEquals(splitPlacementResult.getAssignments().keySet().size(), 2);
    // adding one more split(3 % 3 = 0) that will fall into different slots, 3 splits will be distributed to 3 nodes
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestAffinitySplitRemote(3)));
    splitPlacementResult = nodeSelector.computeAssignments(splits, getRemoteTableScanTask(splitPlacementResult));
    assertEquals(splitPlacementResult.getAssignments().keySet().size(), 3);
}

Example 3 with SplitPlacementResult

use of com.facebook.presto.execution.scheduler.SplitPlacementResult in project presto by prestodb.

the class TestNodeScheduler method testAffinityAssignmentWithConsistentHashing.

@Test
public void testAffinityAssignmentWithConsistentHashing() {
    NodeTaskMap nodeTaskMap = new NodeTaskMap(finalizerService);
    TestingTransactionHandle transactionHandle = TestingTransactionHandle.create();
    NodeSchedulerConfig nodeSchedulerConfig = new NodeSchedulerConfig().setNodeSelectionHashStrategy(CONSISTENT_HASHING).setMinVirtualNodeCount(3).setMaxSplitsPerNode(20).setIncludeCoordinator(false).setMaxPendingSplitsPerTask(10);
    NodeScheduler nodeScheduler = new NodeScheduler(new LegacyNetworkTopology(), nodeManager, new NodeSelectionStats(), nodeSchedulerConfig, nodeTaskMap, new ThrowingNodeTtlFetcherManager(), new NoOpQueryManager(), new SimpleTtlNodeSelectorConfig());
    NodeSelector nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 3);
    Set<Split> splits = new HashSet<>();
    IntStream.range(0, 10).forEach(i -> splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestAffinitySplitRemote(i))));
    InternalNode node1 = new InternalNode("other1", URI.create("http://127.0.0.1:11"), NodeVersion.UNKNOWN, false);
    InternalNode node2 = new InternalNode("other2", URI.create("http://127.0.0.1:12"), NodeVersion.UNKNOWN, false);
    InternalNode node3 = new InternalNode("other3", URI.create("http://127.0.0.1:13"), NodeVersion.UNKNOWN, false);
    InternalNode node4 = new InternalNode("other4", URI.create("http://127.0.0.1:14"), NodeVersion.UNKNOWN, false);
    // In setup node 1-3 are added to node manager
    SplitPlacementResult splitPlacementResult = nodeSelector.computeAssignments(splits, ImmutableList.of());
    assertEquals(splitPlacementResult.getAssignments().keySet().size(), 3);
    // node1: split 1, 3, 4, 5, 6, 7, 8, 9
    Collection<ConnectorSplit> node1Splits = splitPlacementResult.getAssignments().get(node1).stream().map(Split::getConnectorSplit).collect(toImmutableSet());
    // node2: 0
    Collection<Object> node2Splits = splitPlacementResult.getAssignments().get(node2).stream().map(Split::getConnectorSplit).collect(toImmutableSet());
    // node3: split 2
    Collection<Object> node3Splits = splitPlacementResult.getAssignments().get(node3).stream().map(Split::getConnectorSplit).collect(toImmutableSet());
    // Scheduling the same splits on the same set of nodes should give the same assignment
    nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 3);
    splitPlacementResult = nodeSelector.computeAssignments(splits, ImmutableList.of());
    assertEquals(splitPlacementResult.getAssignments().get(node1).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node1Splits);
    assertEquals(splitPlacementResult.getAssignments().get(node2).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node2Splits);
    assertEquals(splitPlacementResult.getAssignments().get(node3).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node3Splits);
    // Adding node4. Node4 is hashed in between node3 and node1, and all splits previously assigned to node1 are now assigned to node4. Assignment to node2 and node3 should not change.
    nodeManager.addNode(CONNECTOR_ID, node4);
    nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 3);
    splitPlacementResult = nodeSelector.computeAssignments(splits, ImmutableList.of());
    assertEquals(splitPlacementResult.getAssignments().keySet().size(), 3);
    assertEquals(splitPlacementResult.getAssignments().get(node1), ImmutableList.of());
    assertEquals(splitPlacementResult.getAssignments().get(node2).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node2Splits);
    assertEquals(splitPlacementResult.getAssignments().get(node3).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node3Splits);
    assertEquals(splitPlacementResult.getAssignments().get(node4).stream().map(Split::getConnectorSplit).collect(toImmutableSet()), node1Splits);
}

Example 4 with SplitPlacementResult

use of com.facebook.presto.execution.scheduler.SplitPlacementResult in project presto by prestodb.

the class TestNodeScheduler method testMaxTasksPerStageAddingNewNodes.

@Test
public void testMaxTasksPerStageAddingNewNodes() {
    InMemoryNodeManager nodeManager = new InMemoryNodeManager();
    NodeTaskMap nodeTaskMap = new NodeTaskMap(finalizerService);
    TestingTransactionHandle transactionHandle = TestingTransactionHandle.create();
    NodeSchedulerConfig nodeSchedulerConfig = new NodeSchedulerConfig().setMaxSplitsPerNode(20).setIncludeCoordinator(false).setMaxPendingSplitsPerTask(10);
    LegacyNetworkTopology networkTopology = new LegacyNetworkTopology();
    // refresh interval is 1 nanosecond
    NodeScheduler nodeScheduler = new NodeScheduler(new NetworkLocationCache(networkTopology), networkTopology, nodeManager, new NodeSelectionStats(), nodeSchedulerConfig, nodeTaskMap, Duration.valueOf("0s"), new ThrowingNodeTtlFetcherManager(), new NoOpQueryManager(), new SimpleTtlNodeSelectorConfig());
    NodeSelector nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 2);
    Set<Split> splits = new HashSet<>();
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestSplitRemote()));
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestSplitRemote()));
    splits.add(new Split(CONNECTOR_ID, transactionHandle, new TestSplitRemote()));
    nodeManager.addNode(CONNECTOR_ID, ImmutableList.of(new InternalNode("node1", URI.create("http://127.0.0.1:11"), NodeVersion.UNKNOWN, false)));
    SplitPlacementResult splitPlacementResult = nodeSelector.computeAssignments(splits, ImmutableList.of());
    Set<InternalNode> internalNodes = splitPlacementResult.getAssignments().keySet();
    assertEquals(internalNodes.size(), 1);
    nodeManager.addNode(CONNECTOR_ID, ImmutableList.of(new InternalNode("node2", URI.create("http://127.0.0.1:12"), NodeVersion.UNKNOWN, false)));
    splitPlacementResult = nodeSelector.computeAssignments(splits, getRemoteTableScanTask(splitPlacementResult));
    Set<InternalNode> internalNodesSecondCall = splitPlacementResult.getAssignments().keySet();
    assertEquals(internalNodesSecondCall.size(), 2);
    assertTrue(internalNodesSecondCall.containsAll(internalNodes));
    nodeManager.addNode(CONNECTOR_ID, ImmutableList.of(new InternalNode("node2", URI.create("http://127.0.0.1:13"), NodeVersion.UNKNOWN, false)));
    internalNodes = splitPlacementResult.getAssignments().keySet();
    assertEquals(internalNodes.size(), 2);
    assertTrue(internalNodesSecondCall.containsAll(internalNodes));
}

Example 5 with SplitPlacementResult

use of com.facebook.presto.execution.scheduler.SplitPlacementResult in project presto by prestodb.

the class TestNodeScheduler method testMoreSplitsAssignedWhenSplitsWeightsAreSmall.

@Test
public void testMoreSplitsAssignedWhenSplitsWeightsAreSmall() {
    int standardSplitsPerNode = nodeSchedulerConfig.getMaxSplitsPerNode();
    int standardPendingSplitsPerTask = nodeSchedulerConfig.getMaxPendingSplitsPerTask();
    int fullyLoadedStandardSplitCount = standardSplitsPerNode + standardPendingSplitsPerTask;
    long weightLimitPerNode = SplitWeight.rawValueForStandardSplitCount(standardSplitsPerNode);
    long weightLimitPendingPerTask = SplitWeight.rawValueForStandardSplitCount(standardPendingSplitsPerTask);
    long fullyLoadedStandardSplitWeight = weightLimitPerNode + weightLimitPendingPerTask;
    // Single worker node
    nodeSelector = nodeScheduler.createNodeSelector(session, CONNECTOR_ID, 1);
    InternalNode workerNode = nodeSelector.selectRandomNodes(1).get(0);
    MockRemoteTaskFactory remoteTaskFactory = new MockRemoteTaskFactory(remoteTaskExecutor, remoteTaskScheduledExecutor);
    TaskId taskId = new TaskId("test", 1, 0, 1);
    MockRemoteTaskFactory.MockRemoteTask task = remoteTaskFactory.createTableScanTask(taskId, workerNode, ImmutableList.of(), nodeTaskMap.createTaskStatsTracker(workerNode, taskId));
    TestingTransactionHandle transactionHandle = TestingTransactionHandle.create();
    ImmutableSet.Builder<Split> splitsBuilder = ImmutableSet.builderWithExpectedSize(fullyLoadedStandardSplitCount * 2);
    // Create 2x more splits than the standard split count limit, at 1/2 the standard weight
    SplitWeight halfWeight = SplitWeight.fromProportion(0.5);
    for (int i = 0; i < fullyLoadedStandardSplitCount * 2; i++) {
        splitsBuilder.add(new Split(CONNECTOR_ID, transactionHandle, new TestSplitRemote(halfWeight)));
    }
    Set<Split> splits = splitsBuilder.build();
    // Verify we arrived at the exact weight limit
    assertEquals(SplitWeight.rawValueSum(splits, Split::getSplitWeight), fullyLoadedStandardSplitWeight);
    // Node assignment limit met
    SplitPlacementResult result = nodeSelector.computeAssignments(splits, ImmutableList.of(task));
    assertEquals(result.getAssignments().get(workerNode).size(), standardSplitsPerNode * 2);
    assertEquals(SplitWeight.rawValueSum(result.getAssignments().get(workerNode), Split::getSplitWeight), weightLimitPerNode);
    // Mark all splits as running
    task.addSplits(ImmutableMultimap.<PlanNodeId, Split>builder().putAll(new PlanNodeId("sourceId"), result.getAssignments().get(workerNode)).build());
    task.startSplits(result.getAssignments().get(workerNode).size());
    // Per task pending splits limit met
    Set<Split> remainingSplits = Sets.difference(splits, ImmutableSet.copyOf(result.getAssignments().get(workerNode)));
    SplitPlacementResult secondResults = nodeSelector.computeAssignments(remainingSplits, ImmutableList.of(task));
    assertEquals(secondResults.getAssignments().get(workerNode).size(), standardPendingSplitsPerTask * 2);
    assertEquals(SplitWeight.rawValueSum(secondResults.getAssignments().get(workerNode), Split::getSplitWeight), weightLimitPendingPerTask);
    task.addSplits(ImmutableMultimap.<PlanNodeId, Split>builder().putAll(new PlanNodeId("sourceId"), secondResults.getAssignments().get(workerNode)).build());
    assertEquals(nodeTaskMap.getPartitionedSplitsOnNode(workerNode), // 2x fully loaded standard count, full weight limit reached
    PartitionedSplitsInfo.forSplitCountAndWeightSum(fullyLoadedStandardSplitCount * 2, fullyLoadedStandardSplitWeight));
    // No more splits assigned when full
    SplitPlacementResult resultWhenFull = nodeSelector.computeAssignments(ImmutableSet.of(new Split(CONNECTOR_ID, transactionHandle, new TestSplitRemote())), ImmutableList.of(task));
    assertTrue(resultWhenFull.getAssignments().isEmpty());
}