/**
* Logs job killed event. Closes the job history log file.
*
* @param timestamp time when job killed was issued in ms.
* @param finishedMaps no finished map tasks.
* @param finishedReduces no of finished reduce tasks.
*/
public void logKilled(long timestamp, int finishedMaps, int finishedReduces, Counters counters) {
if (disableHistory) {
return;
}
if (null != writers) {
log(
writers,
RecordTypes.Job,
new Keys[] {
Keys.JOBID,
Keys.FINISH_TIME,
Keys.JOB_STATUS,
Keys.FINISHED_MAPS,
Keys.FINISHED_REDUCES,
Keys.COUNTERS
},
new String[] {
jobId.toString(),
String.valueOf(timestamp),
Values.KILLED.name(),
String.valueOf(finishedMaps),
String.valueOf(finishedReduces),
counters.makeEscapedCompactString()
},
true);
closeAndClear(writers);
}
}
private File localizeJob(JobID jobid) throws IOException {
String user = UserGroupInformation.getCurrentUser().getShortUserName();
new JobLocalizer(tt.getJobConf(), user, jobid.toString()).initializeJobLogDir();
File jobUserlog = TaskLog.getJobDir(jobid);
JobConf conf = new JobConf();
// localize job log directory
tt.saveLogDir(jobid, conf);
assertTrue(jobUserlog + " directory is not created.", jobUserlog.exists());
return jobUserlog;
}
/**
* Log job finished. closes the job file in history.
*
* @param finishTime finish time of job in ms.
* @param finishedMaps no of maps successfully finished.
* @param finishedReduces no of reduces finished sucessfully.
* @param failedMaps no of failed map tasks. (includes killed)
* @param failedReduces no of failed reduce tasks. (includes killed)
* @param killedMaps no of killed map tasks.
* @param killedReduces no of killed reduce tasks.
* @param counters the counters from the job
*/
public void logFinished(
long finishTime,
int finishedMaps,
int finishedReduces,
int failedMaps,
int failedReduces,
int killedMaps,
int killedReduces,
Counters mapCounters,
Counters reduceCounters,
Counters counters) {
if (disableHistory) {
return;
}
if (null != writers) {
log(
writers,
RecordTypes.Job,
new Keys[] {
Keys.JOBID,
Keys.FINISH_TIME,
Keys.JOB_STATUS,
Keys.FINISHED_MAPS,
Keys.FINISHED_REDUCES,
Keys.FAILED_MAPS,
Keys.FAILED_REDUCES,
Keys.KILLED_MAPS,
Keys.KILLED_REDUCES,
Keys.MAP_COUNTERS,
Keys.REDUCE_COUNTERS,
Keys.COUNTERS
},
new String[] {
jobId.toString(),
Long.toString(finishTime),
Values.SUCCESS.name(),
String.valueOf(finishedMaps),
String.valueOf(finishedReduces),
String.valueOf(failedMaps),
String.valueOf(failedReduces),
String.valueOf(killedMaps),
String.valueOf(killedReduces),
mapCounters.makeEscapedCompactString(),
reduceCounters.makeEscapedCompactString(),
counters.makeEscapedCompactString()
},
true);
closeAndClear(writers);
}
// NOTE: history cleaning stuff deleted from here. We should do that
// somewhere else!
}
/**
* Log job's priority.
*
* @param priority Jobs priority
*/
public void logJobPriority(JobID jobid, JobPriority priority) {
if (disableHistory) {
return;
}
if (null != writers) {
log(
writers,
RecordTypes.Job,
new Keys[] {Keys.JOBID, Keys.JOB_PRIORITY},
new String[] {jobId.toString(), priority.toString()});
}
}
/** Logs job as running */
public void logStarted() {
if (disableHistory) {
return;
}
if (null != writers) {
log(
writers,
RecordTypes.Job,
new Keys[] {Keys.JOBID, Keys.JOB_STATUS},
new String[] {jobId.toString(), Values.RUNNING.name()});
}
}
/**
* Logs launch time of job.
*
* @param startTime start time of job.
* @param totalMaps total maps assigned by jobtracker.
* @param totalReduces total reduces.
*/
public void logInited(long startTime, int totalMaps, int totalReduces) {
if (disableHistory) {
return;
}
if (null != writers) {
log(
writers,
RecordTypes.Job,
new Keys[] {
Keys.JOBID, Keys.LAUNCH_TIME, Keys.TOTAL_MAPS, Keys.TOTAL_REDUCES, Keys.JOB_STATUS
},
new String[] {
jobId.toString(),
String.valueOf(startTime),
String.valueOf(totalMaps),
String.valueOf(totalReduces),
Values.PREP.name()
});
}
}
@SuppressWarnings("deprecation")
public SimulatorJobInProgress(
JobID jobid, JobTracker jobtracker, JobConf default_conf, JobStory jobStory) {
super(jobid, jobStory.getJobConf(), jobtracker);
// jobSetupCleanupNeeded set to false in parent cstr, though
// default is true
restartCount = 0;
jobSetupCleanupNeeded = false;
this.memoryPerMap = conf.getMemoryForMapTask();
this.memoryPerReduce = conf.getMemoryForReduceTask();
this.maxTaskFailuresPerTracker = conf.getMaxTaskFailuresPerTracker();
this.jobId = jobid;
String url =
"http://"
+ jobtracker.getJobTrackerMachine()
+ ":"
+ jobtracker.getInfoPort()
+ "/jobdetails.jsp?jobid="
+ jobid;
this.jobtracker = jobtracker;
this.conf = jobStory.getJobConf();
this.priority = conf.getJobPriority();
Path jobDir = jobtracker.getSystemDirectoryForJob(jobid);
this.jobFile = new Path(jobDir, "job.xml");
this.status =
new JobStatus(jobid, 0.0f, 0.0f, 0.0f, 0.0f, JobStatus.PREP, priority, conf.getUser());
this.profile =
new JobProfile(
jobStory.getUser(),
jobid,
this.jobFile.toString(),
url,
jobStory.getName(),
conf.getQueueName());
this.startTime = JobTracker.getClock().getTime();
status.setStartTime(startTime);
this.resourceEstimator = new ResourceEstimator(this);
this.numMapTasks = jobStory.getNumberMaps();
this.numReduceTasks = jobStory.getNumberReduces();
this.taskCompletionEvents =
new ArrayList<TaskCompletionEvent>(numMapTasks + numReduceTasks + 10);
this.mapFailuresPercent = conf.getMaxMapTaskFailuresPercent();
this.reduceFailuresPercent = conf.getMaxReduceTaskFailuresPercent();
MetricsContext metricsContext = MetricsUtil.getContext("mapred");
this.jobMetrics = MetricsUtil.createRecord(metricsContext, "job");
this.jobMetrics.setTag("user", conf.getUser());
this.jobMetrics.setTag("sessionId", conf.getSessionId());
this.jobMetrics.setTag("jobName", conf.getJobName());
this.jobMetrics.setTag("jobId", jobid.toString());
this.maxLevel = jobtracker.getNumTaskCacheLevels();
this.anyCacheLevel = this.maxLevel + 1;
this.nonLocalMaps = new LinkedList<TaskInProgress>();
this.nonLocalRunningMaps = new LinkedHashSet<TaskInProgress>();
this.runningMapCache = new IdentityHashMap<Node, Set<TaskInProgress>>();
this.nonRunningReduces = new LinkedList<TaskInProgress>();
this.runningReduces = new LinkedHashSet<TaskInProgress>();
this.slowTaskThreshold =
Math.max(0.0f, conf.getFloat("mapred.speculative.execution.slowTaskThreshold", 1.0f));
this.speculativeCap = conf.getFloat("mapred.speculative.execution.speculativeCap", 0.1f);
this.slowNodeThreshold = conf.getFloat("mapred.speculative.execution.slowNodeThreshold", 1.0f);
this.jobStory = jobStory;
// this.jobHistory = this.jobtracker.getJobHistory();
}
/**
* Log job submitted event to history. Creates a new file in history for the job. if history file
* creation fails, it disables history for all other events.
*
* @param jobConfPath path to job conf xml file in HDFS.
* @param submitTime time when job tracker received the job
* @throws IOException
*/
public void logSubmitted(String jobConfPath, long submitTime, String jobTrackerId)
throws IOException {
if (disableHistory) {
return;
}
// create output stream for logging in hadoop.job.history.location
int defaultBufferSize = logDirFs.getConf().getInt("io.file.buffer.size", 4096);
try {
FSDataOutputStream out = null;
PrintWriter writer = null;
// In case the old JT is still running, but we can't connect to it, we
// should ensure that it won't write to our (new JT's) job history file.
if (logDirFs.exists(logFile)) {
LOG.info("Remove the old history file " + logFile);
logDirFs.delete(logFile, true);
}
out =
logDirFs.create(
logFile,
new FsPermission(HISTORY_FILE_PERMISSION),
true,
defaultBufferSize,
logDirFs.getDefaultReplication(),
jobHistoryBlockSize,
null);
writer = new PrintWriter(out);
fileManager.addWriter(jobId, writer);
// cache it ...
fileManager.setHistoryFile(jobId, logFile);
writers = fileManager.getWriters(jobId);
if (null != writers) {
log(
writers,
RecordTypes.Meta,
new Keys[] {Keys.VERSION},
new String[] {String.valueOf(JobHistory.VERSION)});
}
String jobName = getJobName();
String user = getUserName();
// add to writer as well
log(
writers,
RecordTypes.Job,
new Keys[] {
Keys.JOBID, Keys.JOBNAME, Keys.USER, Keys.SUBMIT_TIME, Keys.JOBCONF, Keys.JOBTRACKERID
},
new String[] {
jobId.toString(), jobName, user,
String.valueOf(submitTime), jobConfPath, jobTrackerId
});
} catch (IOException e) {
// Disable history if we have errors other than in the user log.
disableHistory = true;
}
/* Storing the job conf on the log dir */
Path jobFilePath = new Path(logDir, CoronaJobHistoryFilesManager.getConfFilename(jobId));
fileManager.setConfFile(jobId, jobFilePath);
FSDataOutputStream jobFileOut = null;
try {
if (!logDirFs.exists(jobFilePath)) {
jobFileOut =
logDirFs.create(
jobFilePath,
new FsPermission(HISTORY_FILE_PERMISSION),
true,
defaultBufferSize,
logDirFs.getDefaultReplication(),
logDirFs.getDefaultBlockSize(),
null);
conf.writeXml(jobFileOut);
jobFileOut.close();
}
} catch (IOException ioe) {
LOG.error("Failed to store job conf in the log dir", ioe);
} finally {
if (jobFileOut != null) {
try {
jobFileOut.close();
} catch (IOException ie) {
LOG.info(
"Failed to close the job configuration file " + StringUtils.stringifyException(ie));
}
}
}
}