Examples with TopologyBuilder org.apache.storm.topology.TopologyBuilder used on opensource projects

Example 16 with TopologyBuilder

use of org.apache.storm.topology.TopologyBuilder in project storm by apache.

the class StreamBuilder method build.

/**
     * Builds a new {@link StormTopology} for the computation expressed
     * via the stream api.
     *
     * @return the storm topology
     */
public StormTopology build() {
    nodeGroupingInfo.clear();
    windowInfo.clear();
    curGroup.clear();
    TopologicalOrderIterator<Node, Edge> iterator = new TopologicalOrderIterator<>(graph, queue());
    TopologyBuilder topologyBuilder = new TopologyBuilder();
    while (iterator.hasNext()) {
        Node node = iterator.next();
        if (node instanceof SpoutNode) {
            addSpout(topologyBuilder, (SpoutNode) node);
        } else if (node instanceof ProcessorNode) {
            handleProcessorNode((ProcessorNode) node, topologyBuilder);
        } else if (node instanceof PartitionNode) {
            updateNodeGroupingInfo((PartitionNode) node);
            processCurGroup(topologyBuilder);
        } else if (node instanceof WindowNode) {
            updateWindowInfo((WindowNode) node);
            processCurGroup(topologyBuilder);
        } else if (node instanceof SinkNode) {
            processCurGroup(topologyBuilder);
            addSink(topologyBuilder, (SinkNode) node);
        }
    }
    processCurGroup(topologyBuilder);
    mayBeAddTsField();
    return topologyBuilder.createTopology();
}

Example 17 with TopologyBuilder

use of org.apache.storm.topology.TopologyBuilder in project storm by apache.

the class TransactionalTopologyBuilder method buildTopologyBuilder.

public TopologyBuilder buildTopologyBuilder() {
    String coordinator = _spoutId + "/coordinator";
    TopologyBuilder builder = new TopologyBuilder();
    SpoutDeclarer declarer = builder.setSpout(coordinator, new TransactionalSpoutCoordinator(_spout));
    for (Map<String, Object> conf : _spoutConfs) {
        declarer.addConfigurations(conf);
    }
    declarer.addConfiguration(Config.TOPOLOGY_TRANSACTIONAL_ID, _id);
    BoltDeclarer emitterDeclarer = builder.setBolt(_spoutId, new CoordinatedBolt(new TransactionalSpoutBatchExecutor(_spout), null, null), _spoutParallelism).allGrouping(coordinator, TransactionalSpoutCoordinator.TRANSACTION_BATCH_STREAM_ID).addConfiguration(Config.TOPOLOGY_TRANSACTIONAL_ID, _id);
    if (_spout instanceof ICommitterTransactionalSpout) {
        emitterDeclarer.allGrouping(coordinator, TransactionalSpoutCoordinator.TRANSACTION_COMMIT_STREAM_ID);
    }
    for (String id : _bolts.keySet()) {
        Component component = _bolts.get(id);
        Map<String, SourceArgs> coordinatedArgs = new HashMap<String, SourceArgs>();
        for (String c : componentBoltSubscriptions(component)) {
            coordinatedArgs.put(c, SourceArgs.all());
        }
        IdStreamSpec idSpec = null;
        if (component.committer) {
            idSpec = IdStreamSpec.makeDetectSpec(coordinator, TransactionalSpoutCoordinator.TRANSACTION_COMMIT_STREAM_ID);
        }
        BoltDeclarer input = builder.setBolt(id, new CoordinatedBolt(component.bolt, coordinatedArgs, idSpec), component.parallelism);
        for (Map conf : component.componentConfs) {
            input.addConfigurations(conf);
        }
        for (String c : componentBoltSubscriptions(component)) {
            input.directGrouping(c, Constants.COORDINATED_STREAM_ID);
        }
        for (InputDeclaration d : component.declarations) {
            d.declare(input);
        }
        if (component.committer) {
            input.allGrouping(coordinator, TransactionalSpoutCoordinator.TRANSACTION_COMMIT_STREAM_ID);
        }
    }
    return builder;
}

Example 18 with TopologyBuilder

use of org.apache.storm.topology.TopologyBuilder in project storm by apache.

the class TestResourceAwareScheduler method testTopologyWorkerMaxHeapSize.

@Test
public void testTopologyWorkerMaxHeapSize() {
    // Test1: If RAS spreads executors across multiple workers based on the set limit for a worker used by the topology
    INimbus iNimbus = new TestUtilsForResourceAwareScheduler.INimbusTest();
    Map<String, Number> resourceMap = new HashMap<>();
    resourceMap.put(Config.SUPERVISOR_CPU_CAPACITY, 400.0);
    resourceMap.put(Config.SUPERVISOR_MEMORY_CAPACITY_MB, 2000.0);
    Map<String, SupervisorDetails> supMap = TestUtilsForResourceAwareScheduler.genSupervisors(2, 2, resourceMap);
    TopologyBuilder builder1 = new TopologyBuilder();
    builder1.setSpout("wordSpout1", new TestWordSpout(), 4);
    StormTopology stormTopology1 = builder1.createTopology();
    Config config1 = new Config();
    config1.putAll(defaultTopologyConf);
    config1.put(Config.TOPOLOGY_WORKER_MAX_HEAP_SIZE_MB, 128.0);
    Map<ExecutorDetails, String> executorMap1 = TestUtilsForResourceAwareScheduler.genExecsAndComps(stormTopology1);
    TopologyDetails topology1 = new TopologyDetails("topology1", config1, stormTopology1, 1, executorMap1, 0);
    Cluster cluster = new Cluster(iNimbus, supMap, new HashMap<String, SchedulerAssignmentImpl>(), config1);
    ResourceAwareScheduler rs = new ResourceAwareScheduler();
    Map<String, TopologyDetails> topoMap = new HashMap<>();
    topoMap.put(topology1.getId(), topology1);
    Topologies topologies = new Topologies(topoMap);
    rs.prepare(config1);
    rs.schedule(topologies, cluster);
    Assert.assertEquals("Running - Fully Scheduled by DefaultResourceAwareStrategy", cluster.getStatusMap().get(topology1.getId()));
    Assert.assertEquals(4, cluster.getAssignedNumWorkers(topology1));
    // Test2: test when no more workers are available due to topology worker max heap size limit but there is memory is still available
    // wordSpout2 is going to contain 5 executors that needs scheduling. Each of those executors has a memory requirement of 128.0 MB
    // The cluster contains 4 free WorkerSlots. For this topolology each worker is limited to a max heap size of 128.0
    // Thus, one executor not going to be able to get scheduled thus failing the scheduling of this topology and no executors of this topology will be scheduleded
    TopologyBuilder builder2 = new TopologyBuilder();
    builder2.setSpout("wordSpout2", new TestWordSpout(), 5);
    StormTopology stormTopology2 = builder2.createTopology();
    Config config2 = new Config();
    config2.putAll(defaultTopologyConf);
    config2.put(Config.TOPOLOGY_WORKER_MAX_HEAP_SIZE_MB, 128.0);
    Map<ExecutorDetails, String> executorMap2 = TestUtilsForResourceAwareScheduler.genExecsAndComps(stormTopology2);
    TopologyDetails topology2 = new TopologyDetails("topology2", config2, stormTopology2, 1, executorMap2, 0);
    cluster = new Cluster(iNimbus, supMap, new HashMap<String, SchedulerAssignmentImpl>(), config2);
    topoMap = new HashMap<>();
    topoMap.put(topology2.getId(), topology2);
    topologies = new Topologies(topoMap);
    rs.prepare(config2);
    rs.schedule(topologies, cluster);
    Assert.assertEquals("Not enough resources to schedule - 0/5 executors scheduled", cluster.getStatusMap().get(topology2.getId()));
    Assert.assertEquals(5, cluster.getUnassignedExecutors(topology2).size());
}

Example 19 with TopologyBuilder

use of org.apache.storm.topology.TopologyBuilder in project storm by apache.

the class TestResourceAwareScheduler method TestMultipleSpoutsAndCyclicTopologies.

/**
     * Test multiple spouts and cyclic topologies
     */
@Test
public void TestMultipleSpoutsAndCyclicTopologies() {
    TopologyBuilder builder = new TopologyBuilder();
    SpoutDeclarer s1 = builder.setSpout("spout-1", new TestUtilsForResourceAwareScheduler.TestSpout(), 5);
    SpoutDeclarer s2 = builder.setSpout("spout-2", new TestUtilsForResourceAwareScheduler.TestSpout(), 5);
    BoltDeclarer b1 = builder.setBolt("bolt-1", new TestUtilsForResourceAwareScheduler.TestBolt(), 5).shuffleGrouping("spout-1").shuffleGrouping("bolt-3");
    BoltDeclarer b2 = builder.setBolt("bolt-2", new TestUtilsForResourceAwareScheduler.TestBolt(), 5).shuffleGrouping("bolt-1");
    BoltDeclarer b3 = builder.setBolt("bolt-3", new TestUtilsForResourceAwareScheduler.TestBolt(), 5).shuffleGrouping("bolt-2").shuffleGrouping("spout-2");
    INimbus iNimbus = new TestUtilsForResourceAwareScheduler.INimbusTest();
    Map<String, Number> resourceMap = new HashMap<String, Number>();
    resourceMap.put(Config.SUPERVISOR_CPU_CAPACITY, 100.0);
    resourceMap.put(Config.SUPERVISOR_MEMORY_CAPACITY_MB, 1000.0);
    Map<String, SupervisorDetails> supMap = TestUtilsForResourceAwareScheduler.genSupervisors(25, 1, resourceMap);
    Config config = new Config();
    config.putAll(Utils.readDefaultConfig());
    config.put(Config.RESOURCE_AWARE_SCHEDULER_EVICTION_STRATEGY, org.apache.storm.scheduler.resource.strategies.eviction.DefaultEvictionStrategy.class.getName());
    config.put(Config.RESOURCE_AWARE_SCHEDULER_PRIORITY_STRATEGY, org.apache.storm.scheduler.resource.strategies.priority.DefaultSchedulingPriorityStrategy.class.getName());
    config.put(Config.TOPOLOGY_SCHEDULER_STRATEGY, org.apache.storm.scheduler.resource.strategies.scheduling.DefaultResourceAwareStrategy.class.getName());
    config.put(Config.TOPOLOGY_COMPONENT_CPU_PCORE_PERCENT, 100.0);
    config.put(Config.TOPOLOGY_COMPONENT_RESOURCES_OFFHEAP_MEMORY_MB, 500);
    config.put(Config.TOPOLOGY_COMPONENT_RESOURCES_ONHEAP_MEMORY_MB, 500);
    config.put(Config.TOPOLOGY_WORKER_MAX_HEAP_SIZE_MB, Double.MAX_VALUE);
    StormTopology stormTopology = builder.createTopology();
    TopologyDetails topo = new TopologyDetails("topo-1", config, stormTopology, 0, genExecsAndComps(stormTopology), 0);
    Cluster cluster = new Cluster(iNimbus, supMap, new HashMap<String, SchedulerAssignmentImpl>(), config);
    config.put(Config.TOPOLOGY_SUBMITTER_USER, "jerry");
    Map<String, TopologyDetails> topoMap = new HashMap<String, TopologyDetails>();
    topoMap.put(topo.getId(), topo);
    Topologies topologies = new Topologies(topoMap);
    ResourceAwareScheduler rs = new ResourceAwareScheduler();
    rs.prepare(config);
    rs.schedule(topologies, cluster);
    Assert.assertTrue("Topo scheduled?", cluster.getAssignmentById(topo.getId()) != null);
    Assert.assertEquals("Topo all executors scheduled?", 25, cluster.getAssignmentById(topo.getId()).getExecutorToSlot().size());
}

Example 20 with TopologyBuilder

use of org.apache.storm.topology.TopologyBuilder in project storm by apache.

the class TestResourceAwareScheduler method testResourceLimitation.

@Test
public void testResourceLimitation() {
    INimbus iNimbus = new TestUtilsForResourceAwareScheduler.INimbusTest();
    Map<String, Number> resourceMap = new HashMap<>();
    resourceMap.put(Config.SUPERVISOR_CPU_CAPACITY, 400.0);
    resourceMap.put(Config.SUPERVISOR_MEMORY_CAPACITY_MB, 2000.0);
    Map<String, SupervisorDetails> supMap = TestUtilsForResourceAwareScheduler.genSupervisors(2, 2, resourceMap);
    // a topology with multiple spouts
    TopologyBuilder builder1 = new TopologyBuilder();
    builder1.setSpout("wordSpout", new TestWordSpout(), 2).setCPULoad(250.0).setMemoryLoad(1000.0, 200.0);
    builder1.setBolt("wordCountBolt", new TestWordCounter(), 1).shuffleGrouping("wordSpout").setCPULoad(100.0).setMemoryLoad(500.0, 100.0);
    StormTopology stormTopology1 = builder1.createTopology();
    Config config = new Config();
    config.putAll(defaultTopologyConf);
    Map<ExecutorDetails, String> executorMap1 = TestUtilsForResourceAwareScheduler.genExecsAndComps(stormTopology1);
    TopologyDetails topology1 = new TopologyDetails("topology1", config, stormTopology1, 2, executorMap1, 0);
    Cluster cluster = new Cluster(iNimbus, supMap, new HashMap<String, SchedulerAssignmentImpl>(), config);
    ResourceAwareScheduler rs = new ResourceAwareScheduler();
    Map<String, TopologyDetails> topoMap = new HashMap<>();
    topoMap.put(topology1.getId(), topology1);
    Topologies topologies = new Topologies(topoMap);
    rs.prepare(config);
    rs.schedule(topologies, cluster);
    SchedulerAssignment assignment1 = cluster.getAssignmentById(topology1.getId());
    Set<WorkerSlot> assignedSlots1 = assignment1.getSlots();
    Set<String> nodesIDs1 = new HashSet<>();
    for (WorkerSlot slot : assignedSlots1) {
        nodesIDs1.add(slot.getNodeId());
    }
    Collection<ExecutorDetails> executors1 = assignment1.getExecutors();
    List<Double> assignedExecutorMemory = new ArrayList<>();
    List<Double> assignedExecutorCpu = new ArrayList<>();
    for (ExecutorDetails executor : executors1) {
        assignedExecutorMemory.add(topology1.getTotalMemReqTask(executor));
        assignedExecutorCpu.add(topology1.getTotalCpuReqTask(executor));
    }
    Collections.sort(assignedExecutorCpu);
    Collections.sort(assignedExecutorMemory);
    Map<ExecutorDetails, SupervisorDetails> executorToSupervisor = new HashMap<>();
    Map<SupervisorDetails, List<ExecutorDetails>> supervisorToExecutors = new HashMap<>();
    Map<Double, Double> cpuAvailableToUsed = new HashMap();
    Map<Double, Double> memoryAvailableToUsed = new HashMap();
    for (Map.Entry<ExecutorDetails, WorkerSlot> entry : assignment1.getExecutorToSlot().entrySet()) {
        executorToSupervisor.put(entry.getKey(), cluster.getSupervisorById(entry.getValue().getNodeId()));
    }
    for (Map.Entry<ExecutorDetails, SupervisorDetails> entry : executorToSupervisor.entrySet()) {
        List<ExecutorDetails> executorsOnSupervisor = supervisorToExecutors.get(entry.getValue());
        if (executorsOnSupervisor == null) {
            executorsOnSupervisor = new ArrayList<>();
            supervisorToExecutors.put(entry.getValue(), executorsOnSupervisor);
        }
        executorsOnSupervisor.add(entry.getKey());
    }
    for (Map.Entry<SupervisorDetails, List<ExecutorDetails>> entry : supervisorToExecutors.entrySet()) {
        Double supervisorTotalCpu = entry.getKey().getTotalCPU();
        Double supervisorTotalMemory = entry.getKey().getTotalMemory();
        Double supervisorUsedCpu = 0.0;
        Double supervisorUsedMemory = 0.0;
        for (ExecutorDetails executor : entry.getValue()) {
            supervisorUsedMemory += topology1.getTotalCpuReqTask(executor);
            supervisorTotalCpu += topology1.getTotalMemReqTask(executor);
        }
        cpuAvailableToUsed.put(supervisorTotalCpu, supervisorUsedCpu);
        memoryAvailableToUsed.put(supervisorTotalMemory, supervisorUsedMemory);
    }
    // executor0 resides one one worker (on one), executor1 and executor2 on another worker (on the other node)
    Assert.assertEquals(2, assignedSlots1.size());
    Assert.assertEquals(2, nodesIDs1.size());
    Assert.assertEquals(3, executors1.size());
    Assert.assertEquals(100.0, assignedExecutorCpu.get(0), 0.001);
    Assert.assertEquals(250.0, assignedExecutorCpu.get(1), 0.001);
    Assert.assertEquals(250.0, assignedExecutorCpu.get(2), 0.001);
    Assert.assertEquals(600.0, assignedExecutorMemory.get(0), 0.001);
    Assert.assertEquals(1200.0, assignedExecutorMemory.get(1), 0.001);
    Assert.assertEquals(1200.0, assignedExecutorMemory.get(2), 0.001);
    for (Map.Entry<Double, Double> entry : memoryAvailableToUsed.entrySet()) {
        Assert.assertTrue(entry.getKey() - entry.getValue() >= 0);
    }
    for (Map.Entry<Double, Double> entry : cpuAvailableToUsed.entrySet()) {
        Assert.assertTrue(entry.getKey() - entry.getValue() >= 0);
    }
    Assert.assertEquals("Running - Fully Scheduled by DefaultResourceAwareStrategy", cluster.getStatusMap().get(topology1.getId()));
}