Search in sources :

Example 1 with TaskState

use of org.apache.hadoop.mapreduce.v2.api.records.TaskState in project hadoop by apache.

the class MRApp method waitForState.

public void waitForState(Task task, TaskState finalState) throws Exception {
    int timeoutSecs = 0;
    TaskReport report = task.getReport();
    while (!finalState.equals(report.getTaskState()) && timeoutSecs++ < 20) {
        System.out.println("Task State for " + task.getID() + " is : " + report.getTaskState() + " Waiting for state : " + finalState + "   progress : " + report.getProgress());
        report = task.getReport();
        Thread.sleep(500);
    }
    System.out.println("Task State is : " + report.getTaskState());
    Assert.assertEquals("Task state is not correct (timedout)", finalState, report.getTaskState());
}
Also used : TaskReport(org.apache.hadoop.mapreduce.v2.api.records.TaskReport)

Example 2 with TaskState

use of org.apache.hadoop.mapreduce.v2.api.records.TaskState in project hadoop by apache.

the class CLI method run.

public int run(String[] argv) throws Exception {
    int exitCode = -1;
    if (argv.length < 1) {
        displayUsage("");
        return exitCode;
    }
    // process arguments
    String cmd = argv[0];
    String submitJobFile = null;
    String jobid = null;
    String taskid = null;
    String historyFileOrJobId = null;
    String historyOutFile = null;
    String historyOutFormat = HistoryViewer.HUMAN_FORMAT;
    String counterGroupName = null;
    String counterName = null;
    JobPriority jp = null;
    String taskType = null;
    String taskState = null;
    int fromEvent = 0;
    int nEvents = 0;
    int jpvalue = 0;
    String configOutFile = null;
    boolean getStatus = false;
    boolean getCounter = false;
    boolean killJob = false;
    boolean listEvents = false;
    boolean viewHistory = false;
    boolean viewAllHistory = false;
    boolean listJobs = false;
    boolean listAllJobs = false;
    boolean listActiveTrackers = false;
    boolean listBlacklistedTrackers = false;
    boolean displayTasks = false;
    boolean killTask = false;
    boolean failTask = false;
    boolean setJobPriority = false;
    boolean logs = false;
    boolean downloadConfig = false;
    if ("-submit".equals(cmd)) {
        if (argv.length != 2) {
            displayUsage(cmd);
            return exitCode;
        }
        submitJobFile = argv[1];
    } else if ("-status".equals(cmd)) {
        if (argv.length != 2) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        getStatus = true;
    } else if ("-counter".equals(cmd)) {
        if (argv.length != 4) {
            displayUsage(cmd);
            return exitCode;
        }
        getCounter = true;
        jobid = argv[1];
        counterGroupName = argv[2];
        counterName = argv[3];
    } else if ("-kill".equals(cmd)) {
        if (argv.length != 2) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        killJob = true;
    } else if ("-set-priority".equals(cmd)) {
        if (argv.length != 3) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        try {
            jp = JobPriority.valueOf(argv[2]);
        } catch (IllegalArgumentException iae) {
            try {
                jpvalue = Integer.parseInt(argv[2]);
            } catch (NumberFormatException ne) {
                LOG.info(ne);
                displayUsage(cmd);
                return exitCode;
            }
        }
        setJobPriority = true;
    } else if ("-events".equals(cmd)) {
        if (argv.length != 4) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        fromEvent = Integer.parseInt(argv[2]);
        nEvents = Integer.parseInt(argv[3]);
        listEvents = true;
    } else if ("-history".equals(cmd)) {
        viewHistory = true;
        if (argv.length < 2 || argv.length > 7) {
            displayUsage(cmd);
            return exitCode;
        }
        // Some arguments are optional while others are not, and some require
        // second arguments.  Due to this, the indexing can vary depending on
        // what's specified and what's left out, as summarized in the below table:
        // [all] <jobHistoryFile|jobId> [-outfile <file>] [-format <human|json>]
        //   1                  2            3       4         5         6
        //   1                  2            3       4
        //   1                  2                              3         4
        //   1                  2
        //                      1            2       3         4         5
        //                      1            2       3
        //                      1                              2         3
        //                      1
        // "all" is optional, but comes first if specified
        int index = 1;
        if ("all".equals(argv[index])) {
            index++;
            viewAllHistory = true;
            if (argv.length == 2) {
                displayUsage(cmd);
                return exitCode;
            }
        }
        // Get the job history file or job id argument
        historyFileOrJobId = argv[index++];
        // "-outfile" is optional, but if specified requires a second argument
        if (argv.length > index + 1 && "-outfile".equals(argv[index])) {
            index++;
            historyOutFile = argv[index++];
        }
        // "-format" is optional, but if specified required a second argument
        if (argv.length > index + 1 && "-format".equals(argv[index])) {
            index++;
            historyOutFormat = argv[index++];
        }
        // Check for any extra arguments that don't belong here
        if (argv.length > index) {
            displayUsage(cmd);
            return exitCode;
        }
    } else if ("-list".equals(cmd)) {
        if (argv.length != 1 && !(argv.length == 2 && "all".equals(argv[1]))) {
            displayUsage(cmd);
            return exitCode;
        }
        if (argv.length == 2 && "all".equals(argv[1])) {
            listAllJobs = true;
        } else {
            listJobs = true;
        }
    } else if ("-kill-task".equals(cmd)) {
        if (argv.length != 2) {
            displayUsage(cmd);
            return exitCode;
        }
        killTask = true;
        taskid = argv[1];
    } else if ("-fail-task".equals(cmd)) {
        if (argv.length != 2) {
            displayUsage(cmd);
            return exitCode;
        }
        failTask = true;
        taskid = argv[1];
    } else if ("-list-active-trackers".equals(cmd)) {
        if (argv.length != 1) {
            displayUsage(cmd);
            return exitCode;
        }
        listActiveTrackers = true;
    } else if ("-list-blacklisted-trackers".equals(cmd)) {
        if (argv.length != 1) {
            displayUsage(cmd);
            return exitCode;
        }
        listBlacklistedTrackers = true;
    } else if ("-list-attempt-ids".equals(cmd)) {
        if (argv.length != 4) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        taskType = argv[2];
        taskState = argv[3];
        displayTasks = true;
        if (!taskTypes.contains(org.apache.hadoop.util.StringUtils.toUpperCase(taskType))) {
            System.out.println("Error: Invalid task-type: " + taskType);
            displayUsage(cmd);
            return exitCode;
        }
        if (!taskStates.contains(org.apache.hadoop.util.StringUtils.toLowerCase(taskState))) {
            System.out.println("Error: Invalid task-state: " + taskState);
            displayUsage(cmd);
            return exitCode;
        }
    } else if ("-logs".equals(cmd)) {
        if (argv.length == 2 || argv.length == 3) {
            logs = true;
            jobid = argv[1];
            if (argv.length == 3) {
                taskid = argv[2];
            } else {
                taskid = null;
            }
        } else {
            displayUsage(cmd);
            return exitCode;
        }
    } else if ("-config".equals(cmd)) {
        downloadConfig = true;
        if (argv.length != 3) {
            displayUsage(cmd);
            return exitCode;
        }
        jobid = argv[1];
        configOutFile = argv[2];
    } else {
        displayUsage(cmd);
        return exitCode;
    }
    // initialize cluster
    cluster = createCluster();
    // Submit the request
    try {
        if (submitJobFile != null) {
            Job job = Job.getInstance(new JobConf(submitJobFile));
            job.submit();
            System.out.println("Created job " + job.getJobID());
            exitCode = 0;
        } else if (getStatus) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                Counters counters = job.getCounters();
                System.out.println();
                System.out.println(job);
                if (counters != null) {
                    System.out.println(counters);
                } else {
                    System.out.println("Counters not available. Job is retired.");
                }
                exitCode = 0;
            }
        } else if (getCounter) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                Counters counters = job.getCounters();
                if (counters == null) {
                    System.out.println("Counters not available for retired job " + jobid);
                    exitCode = -1;
                } else {
                    System.out.println(getCounter(counters, counterGroupName, counterName));
                    exitCode = 0;
                }
            }
        } else if (killJob) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                JobStatus jobStatus = job.getStatus();
                if (jobStatus.getState() == JobStatus.State.FAILED) {
                    System.out.println("Could not mark the job " + jobid + " as killed, as it has already failed.");
                    exitCode = -1;
                } else if (jobStatus.getState() == JobStatus.State.KILLED) {
                    System.out.println("The job " + jobid + " has already been killed.");
                    exitCode = -1;
                } else if (jobStatus.getState() == JobStatus.State.SUCCEEDED) {
                    System.out.println("Could not kill the job " + jobid + ", as it has already succeeded.");
                    exitCode = -1;
                } else {
                    job.killJob();
                    System.out.println("Killed job " + jobid);
                    exitCode = 0;
                }
            }
        } else if (setJobPriority) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                if (jp != null) {
                    job.setPriority(jp);
                } else {
                    job.setPriorityAsInteger(jpvalue);
                }
                System.out.println("Changed job priority.");
                exitCode = 0;
            }
        } else if (viewHistory) {
            // it's a Job ID
            if (historyFileOrJobId.endsWith(".jhist")) {
                viewHistory(historyFileOrJobId, viewAllHistory, historyOutFile, historyOutFormat);
                exitCode = 0;
            } else {
                Job job = getJob(JobID.forName(historyFileOrJobId));
                if (job == null) {
                    System.out.println("Could not find job " + jobid);
                } else {
                    String historyUrl = job.getHistoryUrl();
                    if (historyUrl == null || historyUrl.isEmpty()) {
                        System.out.println("History file for job " + historyFileOrJobId + " is currently unavailable.");
                    } else {
                        viewHistory(historyUrl, viewAllHistory, historyOutFile, historyOutFormat);
                        exitCode = 0;
                    }
                }
            }
        } else if (listEvents) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                listEvents(job, fromEvent, nEvents);
                exitCode = 0;
            }
        } else if (listJobs) {
            listJobs(cluster);
            exitCode = 0;
        } else if (listAllJobs) {
            listAllJobs(cluster);
            exitCode = 0;
        } else if (listActiveTrackers) {
            listActiveTrackers(cluster);
            exitCode = 0;
        } else if (listBlacklistedTrackers) {
            listBlacklistedTrackers(cluster);
            exitCode = 0;
        } else if (displayTasks) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                displayTasks(getJob(JobID.forName(jobid)), taskType, taskState);
                exitCode = 0;
            }
        } else if (killTask) {
            TaskAttemptID taskID = TaskAttemptID.forName(taskid);
            Job job = getJob(taskID.getJobID());
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else if (job.killTask(taskID, false)) {
                System.out.println("Killed task " + taskid);
                exitCode = 0;
            } else {
                System.out.println("Could not kill task " + taskid);
                exitCode = -1;
            }
        } else if (failTask) {
            TaskAttemptID taskID = TaskAttemptID.forName(taskid);
            Job job = getJob(taskID.getJobID());
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else if (job.killTask(taskID, true)) {
                System.out.println("Killed task " + taskID + " by failing it");
                exitCode = 0;
            } else {
                System.out.println("Could not fail task " + taskid);
                exitCode = -1;
            }
        } else if (logs) {
            JobID jobID = JobID.forName(jobid);
            if (getJob(jobID) == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                try {
                    TaskAttemptID taskAttemptID = TaskAttemptID.forName(taskid);
                    LogParams logParams = cluster.getLogParams(jobID, taskAttemptID);
                    LogCLIHelpers logDumper = new LogCLIHelpers();
                    logDumper.setConf(getConf());
                    exitCode = logDumper.dumpAContainersLogs(logParams.getApplicationId(), logParams.getContainerId(), logParams.getNodeId(), logParams.getOwner());
                } catch (IOException e) {
                    if (e instanceof RemoteException) {
                        throw e;
                    }
                    System.out.println(e.getMessage());
                }
            }
        } else if (downloadConfig) {
            Job job = getJob(JobID.forName(jobid));
            if (job == null) {
                System.out.println("Could not find job " + jobid);
            } else {
                String jobFile = job.getJobFile();
                if (jobFile == null || jobFile.isEmpty()) {
                    System.out.println("Config file for job " + jobFile + " could not be found.");
                } else {
                    Path configPath = new Path(jobFile);
                    FileSystem fs = FileSystem.get(getConf());
                    fs.copyToLocalFile(configPath, new Path(configOutFile));
                    exitCode = 0;
                }
            }
        }
    } catch (RemoteException re) {
        IOException unwrappedException = re.unwrapRemoteException();
        if (unwrappedException instanceof AccessControlException) {
            System.out.println(unwrappedException.getMessage());
        } else {
            throw re;
        }
    } finally {
        cluster.close();
    }
    return exitCode;
}
Also used : Path(org.apache.hadoop.fs.Path) TaskAttemptID(org.apache.hadoop.mapreduce.TaskAttemptID) JobPriority(org.apache.hadoop.mapreduce.JobPriority) AccessControlException(org.apache.hadoop.security.AccessControlException) IOException(java.io.IOException) LogParams(org.apache.hadoop.mapreduce.v2.LogParams) JobStatus(org.apache.hadoop.mapreduce.JobStatus) FileSystem(org.apache.hadoop.fs.FileSystem) LogCLIHelpers(org.apache.hadoop.yarn.logaggregation.LogCLIHelpers) Counters(org.apache.hadoop.mapreduce.Counters) Job(org.apache.hadoop.mapreduce.Job) RemoteException(org.apache.hadoop.ipc.RemoteException) JobConf(org.apache.hadoop.mapred.JobConf) JobID(org.apache.hadoop.mapreduce.JobID)

Example 3 with TaskState

use of org.apache.hadoop.mapreduce.v2.api.records.TaskState in project hadoop by apache.

the class TestTaskImpl method testScheduleTask.

@Test
public /**
   * {@link TaskState#NEW}->{@link TaskState#SCHEDULED}
   */
void testScheduleTask() {
    LOG.info("--- START: testScheduleTask ---");
    mockTask = createMockTask(TaskType.MAP);
    TaskId taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
}
Also used : TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) Test(org.junit.Test)

Example 4 with TaskState

use of org.apache.hadoop.mapreduce.v2.api.records.TaskState in project hadoop by apache.

the class TaskImpl method recover.

/**
   * Recover a completed task from a previous application attempt
   * @param taskInfo recovered info about the task
   * @param recoverTaskOutput whether to recover task outputs
   * @return state of the task after recovery
   */
private TaskStateInternal recover(TaskInfo taskInfo, OutputCommitter committer, boolean recoverTaskOutput) {
    LOG.info("Recovering task " + taskId + " from prior app attempt, status was " + taskInfo.getTaskStatus());
    scheduledTime = taskInfo.getStartTime();
    sendTaskStartedEvent();
    Collection<TaskAttemptInfo> attemptInfos = taskInfo.getAllTaskAttempts().values();
    if (attemptInfos.size() > 0) {
        metrics.launchedTask(this);
    }
    // recover the attempts for this task in the order they finished
    // so task attempt completion events are ordered properly
    int savedNextAttemptNumber = nextAttemptNumber;
    ArrayList<TaskAttemptInfo> taInfos = new ArrayList<TaskAttemptInfo>(taskInfo.getAllTaskAttempts().values());
    Collections.sort(taInfos, TA_INFO_COMPARATOR);
    for (TaskAttemptInfo taInfo : taInfos) {
        nextAttemptNumber = taInfo.getAttemptId().getId();
        TaskAttemptImpl attempt = addAttempt(Avataar.VIRGIN);
        // handle the recovery inline so attempts complete before task does
        attempt.handle(new TaskAttemptRecoverEvent(attempt.getID(), taInfo, committer, recoverTaskOutput));
        finishedAttempts.add(attempt.getID());
        TaskAttemptCompletionEventStatus taces = null;
        TaskAttemptState attemptState = attempt.getState();
        switch(attemptState) {
            case FAILED:
                taces = TaskAttemptCompletionEventStatus.FAILED;
                break;
            case KILLED:
                taces = TaskAttemptCompletionEventStatus.KILLED;
                break;
            case SUCCEEDED:
                taces = TaskAttemptCompletionEventStatus.SUCCEEDED;
                break;
            default:
                throw new IllegalStateException("Unexpected attempt state during recovery: " + attemptState);
        }
        if (attemptState == TaskAttemptState.FAILED) {
            failedAttempts.add(attempt.getID());
            if (failedAttempts.size() >= maxAttempts) {
                taces = TaskAttemptCompletionEventStatus.TIPFAILED;
            }
        }
        // TODO: this shouldn't be necessary after MAPREDUCE-4330
        if (successfulAttempt == null) {
            handleTaskAttemptCompletion(attempt.getID(), taces);
            if (attemptState == TaskAttemptState.SUCCEEDED) {
                successfulAttempt = attempt.getID();
            }
        }
    }
    nextAttemptNumber = savedNextAttemptNumber;
    TaskStateInternal taskState = TaskStateInternal.valueOf(taskInfo.getTaskStatus());
    switch(taskState) {
        case SUCCEEDED:
            if (successfulAttempt != null) {
                sendTaskSucceededEvents();
            } else {
                LOG.info("Missing successful attempt for task " + taskId + ", recovering as RUNNING");
                // there must have been a fetch failure and the retry wasn't complete
                taskState = TaskStateInternal.RUNNING;
                metrics.runningTask(this);
                addAndScheduleAttempt(Avataar.VIRGIN);
            }
            break;
        case FAILED:
        case KILLED:
            {
                if (taskState == TaskStateInternal.KILLED && attemptInfos.size() == 0) {
                    metrics.endWaitingTask(this);
                }
                TaskFailedEvent tfe = new TaskFailedEvent(taskInfo.getTaskId(), taskInfo.getFinishTime(), taskInfo.getTaskType(), taskInfo.getError(), taskInfo.getTaskStatus(), taskInfo.getFailedDueToAttemptId(), taskInfo.getCounters());
                eventHandler.handle(new JobHistoryEvent(taskId.getJobId(), tfe));
                eventHandler.handle(new JobTaskEvent(taskId, getExternalState(taskState)));
                break;
            }
        default:
            throw new java.lang.AssertionError("Unexpected recovered task state: " + taskState);
    }
    return taskState;
}
Also used : TaskStateInternal(org.apache.hadoop.mapreduce.v2.app.job.TaskStateInternal) ArrayList(java.util.ArrayList) TaskAttemptCompletionEventStatus(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptCompletionEventStatus) JobHistoryEvent(org.apache.hadoop.mapreduce.jobhistory.JobHistoryEvent) TaskAttemptState(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptState) JobTaskEvent(org.apache.hadoop.mapreduce.v2.app.job.event.JobTaskEvent) TaskAttemptInfo(org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo) TaskFailedEvent(org.apache.hadoop.mapreduce.jobhistory.TaskFailedEvent) TaskAttemptRecoverEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptRecoverEvent)

Example 5 with TaskState

use of org.apache.hadoop.mapreduce.v2.api.records.TaskState in project hadoop by apache.

the class MockJobs method newTask.

public static Task newTask(JobId jid, int i, int m, final boolean hasFailedTasks) {
    final TaskId tid = Records.newRecord(TaskId.class);
    tid.setJobId(jid);
    tid.setId(i);
    tid.setTaskType(TASK_TYPES.next());
    final TaskReport report = newTaskReport(tid);
    final Map<TaskAttemptId, TaskAttempt> attempts = newTaskAttempts(tid, m);
    return new Task() {

        @Override
        public TaskId getID() {
            return tid;
        }

        @Override
        public TaskReport getReport() {
            return report;
        }

        @Override
        public Counters getCounters() {
            if (hasFailedTasks) {
                return null;
            }
            return new Counters(TypeConverter.fromYarn(report.getCounters()));
        }

        @Override
        public float getProgress() {
            return report.getProgress();
        }

        @Override
        public TaskType getType() {
            return tid.getTaskType();
        }

        @Override
        public Map<TaskAttemptId, TaskAttempt> getAttempts() {
            return attempts;
        }

        @Override
        public TaskAttempt getAttempt(TaskAttemptId attemptID) {
            return attempts.get(attemptID);
        }

        @Override
        public boolean isFinished() {
            switch(report.getTaskState()) {
                case SUCCEEDED:
                case KILLED:
                case FAILED:
                    return true;
            }
            return false;
        }

        @Override
        public boolean canCommit(TaskAttemptId taskAttemptID) {
            return false;
        }

        @Override
        public TaskState getState() {
            return report.getTaskState();
        }
    };
}
Also used : Task(org.apache.hadoop.mapreduce.v2.app.job.Task) TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskReport(org.apache.hadoop.mapreduce.v2.api.records.TaskReport) TaskAttemptId(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId) Counters(org.apache.hadoop.mapreduce.Counters) TaskAttempt(org.apache.hadoop.mapreduce.v2.app.job.TaskAttempt)

Aggregations

TaskId (org.apache.hadoop.mapreduce.v2.api.records.TaskId)7 Test (org.junit.Test)6 Counters (org.apache.hadoop.mapreduce.Counters)2 TaskAttemptID (org.apache.hadoop.mapreduce.TaskAttemptID)2 JobHistoryEvent (org.apache.hadoop.mapreduce.jobhistory.JobHistoryEvent)2 TaskAttemptId (org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId)2 TaskReport (org.apache.hadoop.mapreduce.v2.api.records.TaskReport)2 TaskAttempt (org.apache.hadoop.mapreduce.v2.app.job.TaskAttempt)2 JobTaskEvent (org.apache.hadoop.mapreduce.v2.app.job.event.JobTaskEvent)2 IOException (java.io.IOException)1 ArrayList (java.util.ArrayList)1 FileSystem (org.apache.hadoop.fs.FileSystem)1 Path (org.apache.hadoop.fs.Path)1 RemoteException (org.apache.hadoop.ipc.RemoteException)1 JobConf (org.apache.hadoop.mapred.JobConf)1 Job (org.apache.hadoop.mapreduce.Job)1 JobID (org.apache.hadoop.mapreduce.JobID)1 JobPriority (org.apache.hadoop.mapreduce.JobPriority)1 JobStatus (org.apache.hadoop.mapreduce.JobStatus)1 Event (org.apache.hadoop.mapreduce.jobhistory.Event)1