Search in sources :

Example 26 with DAGSubmittedEvent

use of org.apache.tez.dag.history.events.DAGSubmittedEvent in project tez by apache.

the class RecoveryService method handle.

public void handle(DAGHistoryEvent event) throws IOException {
    if (stopped.get()) {
        LOG.warn("Igoring event as service stopped, eventType" + event.getHistoryEvent().getEventType());
        return;
    }
    HistoryEventType eventType = event.getHistoryEvent().getEventType();
    if (recoveryFatalErrorOccurred.get()) {
        return;
    }
    if (!started.get()) {
        LOG.warn("Adding event of type " + eventType + " to queue as service not started");
        addToEventQueue(event);
        return;
    }
    TezDAGID dagId = event.getDagID();
    if (eventType.equals(HistoryEventType.DAG_SUBMITTED)) {
        DAGSubmittedEvent dagSubmittedEvent = (DAGSubmittedEvent) event.getHistoryEvent();
        String dagName = dagSubmittedEvent.getDAGName();
        if (dagName != null && dagName.startsWith(TezConstants.TEZ_PREWARM_DAG_NAME_PREFIX)) {
            // Skip recording pre-warm DAG events
            skippedDAGs.add(dagId);
            return;
        }
    }
    if (dagId == null || skippedDAGs.contains(dagId)) {
        if (LOG.isDebugEnabled()) {
            LOG.debug("Skipping event for DAG" + ", eventType=" + eventType + ", dagId=" + (dagId == null ? "null" : dagId.toString()) + ", isSkippedDAG=" + (dagId == null ? "null" : skippedDAGs.contains(dagId)));
        }
        return;
    }
    if (event.getHistoryEvent() instanceof SummaryEvent) {
        synchronized (lock) {
            if (stopped.get()) {
                LOG.warn("Igoring event as service stopped, eventType" + event.getHistoryEvent().getEventType());
                return;
            }
            try {
                SummaryEvent summaryEvent = (SummaryEvent) event.getHistoryEvent();
                handleSummaryEvent(dagId, eventType, summaryEvent);
                if (summaryEvent.writeToRecoveryImmediately()) {
                    handleRecoveryEvent(event);
                    // outputStream may already be closed and removed
                    if (outputStreamMap.containsKey(event.getDagID())) {
                        doFlush(outputStreamMap.get(event.getDagID()), appContext.getClock().getTime());
                    }
                } else {
                    if (LOG.isDebugEnabled()) {
                        LOG.debug("Queueing Non-immediate Summary/Recovery event of type" + eventType.name());
                    }
                    addToEventQueue(event);
                }
                if (eventType.equals(HistoryEventType.DAG_FINISHED)) {
                    LOG.info("DAG completed" + ", dagId=" + event.getDagID() + ", queueSize=" + eventQueue.size());
                    completedDAGs.add(dagId);
                    if (outputStreamMap.containsKey(dagId)) {
                        try {
                            outputStreamMap.get(dagId).close();
                            outputStreamMap.remove(dagId);
                        } catch (IOException ioe) {
                            LOG.warn("Error when trying to flush/close recovery file for" + " dag, dagId=" + event.getDagID());
                        }
                    }
                }
            } catch (IOException ioe) {
                LOG.error("Error handling summary event" + ", eventType=" + event.getHistoryEvent().getEventType(), ioe);
                createFatalErrorFlagDir();
                if (eventType.equals(HistoryEventType.DAG_SUBMITTED)) {
                    // Throw error to tell client that dag submission failed
                    throw ioe;
                }
            }
        }
    } else {
        // All other events just get queued
        if (LOG.isDebugEnabled()) {
            LOG.debug("Queueing Non-Summary Recovery event of type " + eventType.name());
        }
        addToEventQueue(event);
    }
}
Also used : TezDAGID(org.apache.tez.dag.records.TezDAGID) SummaryEvent(org.apache.tez.dag.history.SummaryEvent) HistoryEventType(org.apache.tez.dag.history.HistoryEventType) IOException(java.io.IOException) DAGSubmittedEvent(org.apache.tez.dag.history.events.DAGSubmittedEvent)

Example 27 with DAGSubmittedEvent

use of org.apache.tez.dag.history.events.DAGSubmittedEvent in project tez by apache.

the class TestATSHistoryWithACLs method testDagLoggingEnabled.

/**
 * use mini cluster to verify data do push to ats when
 * the dag logging flag in dagsubmitted event is set on
 * @throws Exception
 */
@Test(timeout = 50000)
public void testDagLoggingEnabled() throws Exception {
    ATSHistoryLoggingService historyLoggingService;
    historyLoggingService = ReflectionUtils.createClazzInstance(ATSHistoryLoggingService.class.getName());
    AppContext appContext = mock(AppContext.class);
    when(appContext.getApplicationID()).thenReturn(ApplicationId.newInstance(0, 1));
    historyLoggingService.setAppContext(appContext);
    TezConfiguration tezConf = new TezConfiguration(mrrTezCluster.getConfig());
    String viewAcls = "nobody nobody_group";
    tezConf.set(TezConfiguration.TEZ_AM_VIEW_ACLS, viewAcls);
    tezConf.set(TezConfiguration.TEZ_HISTORY_LOGGING_SERVICE_CLASS, ATSHistoryLoggingService.class.getName());
    Path remoteStagingDir = remoteFs.makeQualified(new Path("/tmp", String.valueOf(random.nextInt(100000))));
    remoteFs.mkdirs(remoteStagingDir);
    tezConf.set(TezConfiguration.TEZ_AM_STAGING_DIR, remoteStagingDir.toString());
    historyLoggingService.init(tezConf);
    historyLoggingService.start();
    ApplicationId appId = ApplicationId.newInstance(100l, 1);
    TezDAGID tezDAGID = TezDAGID.getInstance(appId, 11);
    ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 1);
    DAGPlan dagPlan = DAGPlan.newBuilder().setName("DAGPlanMock").build();
    DAGSubmittedEvent submittedEvent = new DAGSubmittedEvent(tezDAGID, 1, dagPlan, appAttemptId, null, "usr", tezConf, null, null);
    submittedEvent.setHistoryLoggingEnabled(true);
    DAGHistoryEvent event = new DAGHistoryEvent(tezDAGID, submittedEvent);
    historyLoggingService.handle(new DAGHistoryEvent(tezDAGID, submittedEvent));
    Thread.sleep(1000l);
    String url = "http://" + timelineAddress + "/ws/v1/timeline/TEZ_DAG_ID/" + event.getDagID();
    Client client = new Client();
    WebResource resource = client.resource(url);
    ClientResponse response = resource.accept(MediaType.APPLICATION_JSON).get(ClientResponse.class);
    assertEquals(200, response.getStatus());
    assertEquals(MediaType.APPLICATION_JSON_TYPE, response.getType());
    TimelineEntity entity = response.getEntity(TimelineEntity.class);
    assertEquals(entity.getEntityType(), "TEZ_DAG_ID");
    assertEquals(entity.getEvents().get(0).getEventType(), HistoryEventType.DAG_SUBMITTED.toString());
}
Also used : Path(org.apache.hadoop.fs.Path) ClientResponse(com.sun.jersey.api.client.ClientResponse) AppContext(org.apache.tez.dag.app.AppContext) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) WebResource(com.sun.jersey.api.client.WebResource) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) TimelineEntity(org.apache.hadoop.yarn.api.records.timeline.TimelineEntity) ATSHistoryLoggingService(org.apache.tez.dag.history.logging.ats.ATSHistoryLoggingService) DAGPlan(org.apache.tez.dag.api.records.DAGProtos.DAGPlan) TezDAGID(org.apache.tez.dag.records.TezDAGID) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) DAGClient(org.apache.tez.dag.api.client.DAGClient) TezClient(org.apache.tez.client.TezClient) Client(com.sun.jersey.api.client.Client) TezConfiguration(org.apache.tez.dag.api.TezConfiguration) DAGSubmittedEvent(org.apache.tez.dag.history.events.DAGSubmittedEvent) Test(org.junit.Test)

Example 28 with DAGSubmittedEvent

use of org.apache.tez.dag.history.events.DAGSubmittedEvent in project tez by apache.

the class RecoveryParser method parseRecoveryData.

/**
 * 1. Read Summary Recovery file and build DAGSummaryData
 *    Check whether it is recoverable based on the summary file (whether dag is
 *    in the middle of committing)
 * 2. Read the non-Summary Recovery file and build DAGRecoveryData
 *    Check whether it is recoverable based on both the summary file and non-summary file
 *    (whether vertex has completed its committing, but its full non-summary recovery events are not seen)
 * @return DAGRecoveryData
 * @throws IOException
 */
public DAGRecoveryData parseRecoveryData() throws IOException {
    int dagCounter = 0;
    Map<TezDAGID, DAGSummaryData> dagSummaryDataMap = new HashMap<TezDAGID, DAGSummaryData>();
    List<Path> summaryFiles = getSummaryFiles();
    LOG.debug("SummaryFile size:" + summaryFiles.size());
    for (Path summaryFile : summaryFiles) {
        FileStatus summaryFileStatus = recoveryFS.getFileStatus(summaryFile);
        LOG.info("Parsing summary file" + ", path=" + summaryFile.toString() + ", len=" + summaryFileStatus.getLen() + ", lastModTime=" + summaryFileStatus.getModificationTime());
        FSDataInputStream summaryStream = getSummaryStream(summaryFile);
        while (true) {
            RecoveryProtos.SummaryEventProto proto;
            try {
                proto = RecoveryProtos.SummaryEventProto.parseDelimitedFrom(summaryStream);
                if (proto == null) {
                    LOG.info("Reached end of summary stream");
                    break;
                }
            } catch (EOFException eof) {
                LOG.info("Reached end of summary stream");
                break;
            }
            HistoryEventType eventType = HistoryEventType.values()[proto.getEventType()];
            if (LOG.isDebugEnabled()) {
                LOG.debug("[RECOVERY SUMMARY]" + " dagId=" + proto.getDagId() + ", timestamp=" + proto.getTimestamp() + ", event=" + eventType);
            }
            TezDAGID dagId;
            try {
                dagId = TezDAGID.fromString(proto.getDagId());
            } catch (IllegalArgumentException e) {
                throw new IOException("Invalid dagId, summary records may be corrupted", e);
            }
            if (dagCounter < dagId.getId()) {
                dagCounter = dagId.getId();
            }
            if (!dagSummaryDataMap.containsKey(dagId)) {
                dagSummaryDataMap.put(dagId, new DAGSummaryData(dagId));
            }
            try {
                dagSummaryDataMap.get(dagId).handleSummaryEvent(proto);
            } catch (Exception e) {
                // any exception when parsing protobuf
                throw new IOException("Error when parsing summary event proto", e);
            }
        }
        summaryStream.close();
    }
    // Set counter for next set of DAGs & update dagNames Set in DAGAppMaster
    dagAppMaster.setDAGCounter(dagCounter);
    for (DAGSummaryData dagSummaryData : dagSummaryDataMap.values()) {
        dagAppMaster.dagIDs.add(dagSummaryData.dagId.toString());
    }
    DAGSummaryData lastInProgressDAGData = getLastCompletedOrInProgressDAG(dagSummaryDataMap);
    if (lastInProgressDAGData == null) {
        LOG.info("Nothing to recover as no uncompleted/completed DAGs found");
        return null;
    }
    TezDAGID lastInProgressDAG = lastInProgressDAGData.dagId;
    if (lastInProgressDAG == null) {
        LOG.info("Nothing to recover as no uncompleted/completed DAGs found");
        return null;
    }
    LOG.info("Checking if DAG is in recoverable state" + ", dagId=" + lastInProgressDAGData.dagId);
    final DAGRecoveryData recoveredDAGData = new DAGRecoveryData(lastInProgressDAGData);
    List<Path> dagRecoveryFiles = getDAGRecoveryFiles(lastInProgressDAG);
    boolean skipAllOtherEvents = false;
    Path lastRecoveryFile = null;
    // to create the DAGImpl)
    for (Path dagRecoveryFile : dagRecoveryFiles) {
        if (skipAllOtherEvents) {
            LOG.warn("Other recovery files will be skipped due to error in the previous recovery file" + lastRecoveryFile);
            break;
        }
        FileStatus fileStatus = recoveryFS.getFileStatus(dagRecoveryFile);
        lastRecoveryFile = dagRecoveryFile;
        LOG.info("Trying to recover dag from recovery file" + ", dagId=" + lastInProgressDAG.toString() + ", dagRecoveryFile=" + dagRecoveryFile + ", len=" + fileStatus.getLen());
        FSDataInputStream dagRecoveryStream = recoveryFS.open(dagRecoveryFile, recoveryBufferSize);
        while (true) {
            HistoryEvent event;
            try {
                event = getNextEvent(dagRecoveryStream);
                if (event == null) {
                    LOG.info("Reached end of dag recovery stream");
                    break;
                }
            } catch (EOFException eof) {
                LOG.info("Reached end of dag recovery stream");
                break;
            } catch (IOException ioe) {
                LOG.warn("Corrupt data found when trying to read next event", ioe);
                break;
            }
            if (skipAllOtherEvents) {
                // hit an error - skip reading other events
                break;
            }
            HistoryEventType eventType = event.getEventType();
            LOG.info("Recovering from event" + ", eventType=" + eventType + ", event=" + event.toString());
            switch(eventType) {
                case DAG_SUBMITTED:
                    DAGSubmittedEvent submittedEvent = (DAGSubmittedEvent) event;
                    recoveredDAGData.recoveredDAG = dagAppMaster.createDAG(submittedEvent.getDAGPlan(), lastInProgressDAG);
                    recoveredDAGData.cumulativeAdditionalResources = submittedEvent.getCumulativeAdditionalLocalResources();
                    recoveredDAGData.recoveredDagID = recoveredDAGData.recoveredDAG.getID();
                    dagAppMaster.setCurrentDAG(recoveredDAGData.recoveredDAG);
                    if (recoveredDAGData.nonRecoverable) {
                        skipAllOtherEvents = true;
                    }
                    break;
                case DAG_INITIALIZED:
                    recoveredDAGData.dagInitedEvent = (DAGInitializedEvent) event;
                    break;
                case DAG_STARTED:
                    recoveredDAGData.dagStartedEvent = (DAGStartedEvent) event;
                    break;
                case DAG_FINISHED:
                    recoveredDAGData.dagFinishedEvent = (DAGFinishedEvent) event;
                    skipAllOtherEvents = true;
                    break;
                case DAG_COMMIT_STARTED:
                case VERTEX_GROUP_COMMIT_STARTED:
                case VERTEX_GROUP_COMMIT_FINISHED:
                case CONTAINER_LAUNCHED:
                    {
                        // Nothing to do for now
                        break;
                    }
                case DAG_KILL_REQUEST:
                    {
                        break;
                    }
                case VERTEX_INITIALIZED:
                    {
                        VertexInitializedEvent vertexInitEvent = (VertexInitializedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.maybeCreateVertexRecoveryData(vertexInitEvent.getVertexID());
                        vertexRecoveryData.vertexInitedEvent = vertexInitEvent;
                        break;
                    }
                case VERTEX_CONFIGURE_DONE:
                    {
                        VertexConfigurationDoneEvent reconfigureDoneEvent = (VertexConfigurationDoneEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.maybeCreateVertexRecoveryData(reconfigureDoneEvent.getVertexID());
                        vertexRecoveryData.vertexConfigurationDoneEvent = reconfigureDoneEvent;
                        break;
                    }
                case VERTEX_STARTED:
                    {
                        VertexStartedEvent vertexStartedEvent = (VertexStartedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.vertexRecoveryDataMap.get(vertexStartedEvent.getVertexID());
                        Preconditions.checkArgument(vertexRecoveryData != null, "No VertexInitializedEvent before VertexStartedEvent");
                        vertexRecoveryData.vertexStartedEvent = vertexStartedEvent;
                        break;
                    }
                case VERTEX_COMMIT_STARTED:
                    {
                        break;
                    }
                case VERTEX_FINISHED:
                    {
                        VertexFinishedEvent vertexFinishedEvent = (VertexFinishedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.maybeCreateVertexRecoveryData(vertexFinishedEvent.getVertexID());
                        vertexRecoveryData.vertexFinishedEvent = vertexFinishedEvent;
                        break;
                    }
                case TASK_STARTED:
                    {
                        TaskStartedEvent taskStartedEvent = (TaskStartedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.vertexRecoveryDataMap.get(taskStartedEvent.getTaskID().getVertexID());
                        Preconditions.checkArgument(vertexRecoveryData != null, "Invalid TaskStartedEvent, its vertex does not exist:" + taskStartedEvent.getTaskID().getVertexID());
                        TaskRecoveryData taskRecoveryData = vertexRecoveryData.maybeCreateTaskRecoveryData(taskStartedEvent.getTaskID());
                        taskRecoveryData.taskStartedEvent = taskStartedEvent;
                        break;
                    }
                case TASK_FINISHED:
                    {
                        TaskFinishedEvent taskFinishedEvent = (TaskFinishedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.vertexRecoveryDataMap.get(taskFinishedEvent.getTaskID().getVertexID());
                        Preconditions.checkArgument(vertexRecoveryData != null, "Invalid TaskFinishedEvent, its vertex does not exist:" + taskFinishedEvent.getTaskID().getVertexID());
                        TaskRecoveryData taskRecoveryData = vertexRecoveryData.maybeCreateTaskRecoveryData(taskFinishedEvent.getTaskID());
                        taskRecoveryData.taskFinishedEvent = taskFinishedEvent;
                        break;
                    }
                case TASK_ATTEMPT_STARTED:
                    {
                        TaskAttemptStartedEvent taStartedEvent = (TaskAttemptStartedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.vertexRecoveryDataMap.get(taStartedEvent.getTaskAttemptID().getTaskID().getVertexID());
                        Preconditions.checkArgument(vertexRecoveryData != null, "Invalid TaskAttemptStartedEvent, its vertexId does not exist, taId=" + taStartedEvent.getTaskAttemptID());
                        TaskRecoveryData taskRecoveryData = vertexRecoveryData.taskRecoveryDataMap.get(taStartedEvent.getTaskAttemptID().getTaskID());
                        Preconditions.checkArgument(taskRecoveryData != null, "Invalid TaskAttemptStartedEvent, its taskId does not exist, taId=" + taStartedEvent.getTaskAttemptID());
                        TaskAttemptRecoveryData taRecoveryData = taskRecoveryData.maybeCreateTaskAttemptRecoveryData(taStartedEvent.getTaskAttemptID());
                        taRecoveryData.taStartedEvent = taStartedEvent;
                        break;
                    }
                case TASK_ATTEMPT_FINISHED:
                    {
                        TaskAttemptFinishedEvent taFinishedEvent = (TaskAttemptFinishedEvent) event;
                        VertexRecoveryData vertexRecoveryData = recoveredDAGData.vertexRecoveryDataMap.get(taFinishedEvent.getTaskAttemptID().getTaskID().getVertexID());
                        Preconditions.checkArgument(vertexRecoveryData != null, "Invalid TaskAttemtFinishedEvent, its vertexId does not exist, taId=" + taFinishedEvent.getTaskAttemptID());
                        TaskRecoveryData taskRecoveryData = vertexRecoveryData.taskRecoveryDataMap.get(taFinishedEvent.getTaskAttemptID().getTaskID());
                        Preconditions.checkArgument(taskRecoveryData != null, "Invalid TaskAttemptFinishedEvent, its taskId does not exist, taId=" + taFinishedEvent.getTaskAttemptID());
                        TaskAttemptRecoveryData taRecoveryData = taskRecoveryData.maybeCreateTaskAttemptRecoveryData(taFinishedEvent.getTaskAttemptID());
                        taRecoveryData.taFinishedEvent = taFinishedEvent;
                        break;
                    }
                default:
                    throw new RuntimeException("Invalid data found, unknown event type " + eventType);
            }
            if (LOG.isDebugEnabled()) {
                LOG.debug("[DAG RECOVERY]" + " dagId=" + lastInProgressDAG + ", eventType=" + eventType + ", event=" + event.toString());
            }
        }
        dagRecoveryStream.close();
    }
    recoveredDAGData.checkRecoverableNonSummary();
    return recoveredDAGData;
}
Also used : FileStatus(org.apache.hadoop.fs.FileStatus) VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) HashMap(java.util.HashMap) SummaryEventProto(org.apache.tez.dag.recovery.records.RecoveryProtos.SummaryEventProto) HistoryEventType(org.apache.tez.dag.history.HistoryEventType) TezDAGID(org.apache.tez.dag.records.TezDAGID) EOFException(java.io.EOFException) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) Path(org.apache.hadoop.fs.Path) VertexStartedEvent(org.apache.tez.dag.history.events.VertexStartedEvent) RecoveryProtos(org.apache.tez.dag.recovery.records.RecoveryProtos) IOException(java.io.IOException) HistoryEvent(org.apache.tez.dag.history.HistoryEvent) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) EOFException(java.io.EOFException) IOException(java.io.IOException) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) FSDataInputStream(org.apache.hadoop.fs.FSDataInputStream) VertexFinishedEvent(org.apache.tez.dag.history.events.VertexFinishedEvent) DAGSubmittedEvent(org.apache.tez.dag.history.events.DAGSubmittedEvent)

Example 29 with DAGSubmittedEvent

use of org.apache.tez.dag.history.events.DAGSubmittedEvent in project tez by apache.

the class RecoveryParser method getNextEvent.

private static HistoryEvent getNextEvent(FSDataInputStream inputStream) throws IOException {
    int eventTypeOrdinal = -1;
    try {
        eventTypeOrdinal = inputStream.readInt();
    } catch (EOFException eof) {
        return null;
    }
    if (eventTypeOrdinal < 0 || eventTypeOrdinal >= HistoryEventType.values().length) {
        // reached end
        throw new IOException("Corrupt data found when trying to read next event type" + ", eventTypeOrdinal=" + eventTypeOrdinal);
    }
    HistoryEventType eventType = HistoryEventType.values()[eventTypeOrdinal];
    HistoryEvent event;
    switch(eventType) {
        case AM_LAUNCHED:
            event = new AMLaunchedEvent();
            break;
        case AM_STARTED:
            event = new AMStartedEvent();
            break;
        case DAG_SUBMITTED:
            event = new DAGSubmittedEvent();
            break;
        case DAG_INITIALIZED:
            event = new DAGInitializedEvent();
            break;
        case DAG_STARTED:
            event = new DAGStartedEvent();
            break;
        case DAG_COMMIT_STARTED:
            event = new DAGCommitStartedEvent();
            break;
        case DAG_FINISHED:
            event = new DAGFinishedEvent();
            break;
        case DAG_KILL_REQUEST:
            event = new DAGKillRequestEvent();
            break;
        case CONTAINER_LAUNCHED:
            event = new ContainerLaunchedEvent();
            break;
        case CONTAINER_STOPPED:
            event = new ContainerStoppedEvent();
            break;
        case VERTEX_INITIALIZED:
            event = new VertexInitializedEvent();
            break;
        case VERTEX_CONFIGURE_DONE:
            event = new VertexConfigurationDoneEvent();
            break;
        case VERTEX_STARTED:
            event = new VertexStartedEvent();
            break;
        case VERTEX_COMMIT_STARTED:
            event = new VertexCommitStartedEvent();
            break;
        case VERTEX_GROUP_COMMIT_STARTED:
            event = new VertexGroupCommitStartedEvent();
            break;
        case VERTEX_GROUP_COMMIT_FINISHED:
            event = new VertexGroupCommitFinishedEvent();
            break;
        case VERTEX_FINISHED:
            event = new VertexFinishedEvent();
            break;
        case TASK_STARTED:
            event = new TaskStartedEvent();
            break;
        case TASK_FINISHED:
            event = new TaskFinishedEvent();
            break;
        case TASK_ATTEMPT_STARTED:
            event = new TaskAttemptStartedEvent();
            break;
        case TASK_ATTEMPT_FINISHED:
            event = new TaskAttemptFinishedEvent();
            break;
        default:
            throw new IOException("Invalid data found, unknown event type " + eventType);
    }
    try {
        event.fromProtoStream(inputStream);
    } catch (EOFException eof) {
        return null;
    }
    if (LOG.isDebugEnabled()) {
        LOG.debug("Parsed event from input stream" + ", eventType=" + eventType + ", event=" + event.toString());
    }
    return event;
}
Also used : DAGCommitStartedEvent(org.apache.tez.dag.history.events.DAGCommitStartedEvent) VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) HistoryEventType(org.apache.tez.dag.history.HistoryEventType) DAGInitializedEvent(org.apache.tez.dag.history.events.DAGInitializedEvent) DAGKillRequestEvent(org.apache.tez.dag.history.events.DAGKillRequestEvent) ContainerStoppedEvent(org.apache.tez.dag.history.events.ContainerStoppedEvent) EOFException(java.io.EOFException) DAGStartedEvent(org.apache.tez.dag.history.events.DAGStartedEvent) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) AMStartedEvent(org.apache.tez.dag.history.events.AMStartedEvent) VertexStartedEvent(org.apache.tez.dag.history.events.VertexStartedEvent) VertexGroupCommitStartedEvent(org.apache.tez.dag.history.events.VertexGroupCommitStartedEvent) IOException(java.io.IOException) HistoryEvent(org.apache.tez.dag.history.HistoryEvent) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) VertexGroupCommitFinishedEvent(org.apache.tez.dag.history.events.VertexGroupCommitFinishedEvent) AMLaunchedEvent(org.apache.tez.dag.history.events.AMLaunchedEvent) ContainerLaunchedEvent(org.apache.tez.dag.history.events.ContainerLaunchedEvent) DAGFinishedEvent(org.apache.tez.dag.history.events.DAGFinishedEvent) VertexFinishedEvent(org.apache.tez.dag.history.events.VertexFinishedEvent) DAGSubmittedEvent(org.apache.tez.dag.history.events.DAGSubmittedEvent) VertexCommitStartedEvent(org.apache.tez.dag.history.events.VertexCommitStartedEvent)

Example 30 with DAGSubmittedEvent

use of org.apache.tez.dag.history.events.DAGSubmittedEvent in project tez by apache.

the class DAGAppMaster method startDAG.

private void startDAG(DAGPlan dagPlan, Map<String, LocalResource> additionalAMResources) throws TezException {
    long submitTime = this.clock.getTime();
    this.appName = dagPlan.getName();
    // /////////////////// Create the job itself.
    final DAG newDAG = createDAG(dagPlan);
    _updateLoggers(newDAG, "");
    if (LOG.isDebugEnabled()) {
        LOG.debug("Running a DAG with " + dagPlan.getVertexCount() + " vertices ");
        for (VertexPlan v : dagPlan.getVertexList()) {
            LOG.debug("DAG has vertex " + v.getName());
        }
    }
    Map<String, LocalResource> lrDiff = getAdditionalLocalResourceDiff(newDAG, additionalAMResources);
    if (lrDiff != null) {
        amResources.putAll(lrDiff);
        cumulativeAdditionalResources.putAll(lrDiff);
    }
    String callerContextStr = "";
    if (dagPlan.hasCallerContext()) {
        CallerContext callerContext = DagTypeConverters.convertCallerContextFromProto(dagPlan.getCallerContext());
        callerContextStr = ", callerContext=" + callerContext.contextAsSimpleString();
    }
    LOG.info("Running DAG: " + dagPlan.getName() + callerContextStr);
    String timeStamp = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(Calendar.getInstance().getTime());
    System.err.println(timeStamp + " Running Dag: " + newDAG.getID());
    System.out.println(timeStamp + " Running Dag: " + newDAG.getID());
    // Job name is the same as the app name until we support multiple dags
    // for an app later
    final DAGSubmittedEvent submittedEvent = new DAGSubmittedEvent(newDAG.getID(), submitTime, dagPlan, this.appAttemptID, cumulativeAdditionalResources, newDAG.getUserName(), newDAG.getConf(), containerLogs, getContext().getQueueName());
    boolean dagLoggingEnabled = newDAG.getConf().getBoolean(TezConfiguration.TEZ_DAG_HISTORY_LOGGING_ENABLED, TezConfiguration.TEZ_DAG_HISTORY_LOGGING_ENABLED_DEFAULT);
    submittedEvent.setHistoryLoggingEnabled(dagLoggingEnabled);
    try {
        appMasterUgi.doAs(new PrivilegedExceptionAction<Void>() {

            @Override
            public Void run() throws Exception {
                historyEventHandler.handleCriticalEvent(new DAGHistoryEvent(newDAG.getID(), submittedEvent));
                return null;
            }
        });
    } catch (IOException e) {
        throw new TezUncheckedException(e);
    } catch (InterruptedException e) {
        throw new TezUncheckedException(e);
    }
    startDAGExecution(newDAG, lrDiff);
    // set state after curDag is set
    this.state = DAGAppMasterState.RUNNING;
}
Also used : CallerContext(org.apache.tez.client.CallerContext) TezUncheckedException(org.apache.tez.dag.api.TezUncheckedException) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) DAG(org.apache.tez.dag.app.dag.DAG) IOException(java.io.IOException) TezUncheckedException(org.apache.tez.dag.api.TezUncheckedException) IOException(java.io.IOException) UnknownHostException(java.net.UnknownHostException) JSONException(org.codehaus.jettison.json.JSONException) URISyntaxException(java.net.URISyntaxException) FileNotFoundException(java.io.FileNotFoundException) TezException(org.apache.tez.dag.api.TezException) ServiceStateException(org.apache.hadoop.service.ServiceStateException) LocalResource(org.apache.hadoop.yarn.api.records.LocalResource) VertexPlan(org.apache.tez.dag.api.records.DAGProtos.VertexPlan) SimpleDateFormat(java.text.SimpleDateFormat) DAGSubmittedEvent(org.apache.tez.dag.history.events.DAGSubmittedEvent)

Aggregations

DAGSubmittedEvent (org.apache.tez.dag.history.events.DAGSubmittedEvent)31 TezDAGID (org.apache.tez.dag.records.TezDAGID)22 Configuration (org.apache.hadoop.conf.Configuration)21 DAGHistoryEvent (org.apache.tez.dag.history.DAGHistoryEvent)21 DAGPlan (org.apache.tez.dag.api.records.DAGProtos.DAGPlan)18 Path (org.apache.hadoop.fs.Path)16 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)15 SystemClock (org.apache.hadoop.yarn.util.SystemClock)13 DAGRecoveryData (org.apache.tez.dag.app.RecoveryParser.DAGRecoveryData)12 RecoveryService (org.apache.tez.dag.history.recovery.RecoveryService)11 TaskAttemptStartedEvent (org.apache.tez.dag.history.events.TaskAttemptStartedEvent)8 TaskStartedEvent (org.apache.tez.dag.history.events.TaskStartedEvent)8 VertexFinishedEvent (org.apache.tez.dag.history.events.VertexFinishedEvent)8 VertexStartedEvent (org.apache.tez.dag.history.events.VertexStartedEvent)8 TezVertexID (org.apache.tez.dag.records.TezVertexID)8 TezConfiguration (org.apache.tez.dag.api.TezConfiguration)7 HistoryEventType (org.apache.tez.dag.history.HistoryEventType)7 DAGCommitStartedEvent (org.apache.tez.dag.history.events.DAGCommitStartedEvent)7 DAGInitializedEvent (org.apache.tez.dag.history.events.DAGInitializedEvent)7 HistoryEvent (org.apache.tez.dag.history.HistoryEvent)6