public HistoryFileInfo addIfAbsent(HistoryFileInfo fileInfo) {
JobId jobId = fileInfo.getJobId();
if (LogGlobal.isDebugEnabled()) {
/* LOG.debug("Adding "+jobId+" to job list cache with "+fileInfo.getJobIndexInfo()) */
LOG.adding_job_list_cache_with(jobId.toString(), String.valueOf(fileInfo.getJobIndexInfo()))
.tag("methodCall")
.debug();
}
HistoryFileInfo old = cache.putIfAbsent(jobId, fileInfo);
if (cache.size() > maxSize) {
// There is a race here, where more then one thread could be trying to
// remove entries. This could result in too many entries being removed
// from the cache. This is considered OK as the size of the cache
// should be rather large, and we would rather have performance over
// keeping the cache size exactly at the maximum.
Iterator<JobId> keys = cache.navigableKeySet().iterator();
long cutoff = System.currentTimeMillis() - maxAge;
while (cache.size() > maxSize && keys.hasNext()) {
JobId key = keys.next();
HistoryFileInfo firstValue = cache.get(key);
if (firstValue != null) {
synchronized (firstValue) {
if (firstValue.isMovePending()) {
if (firstValue.didMoveFail() && firstValue.jobIndexInfo.getFinishTime() <= cutoff) {
cache.remove(key);
// Now lets try to delete it
try {
firstValue.delete();
} catch (IOException e) {
/* LOG.error("Error while trying to delete history files"+" that could not be moved to done.",e) */
LOG.error_while_trying_delete_history_files_(e.toString()).error();
}
} else {
/* LOG.warn("Waiting to remove "+key+" from JobListCache because it is not in done yet.") */
LOG.waiting_remove_from_joblistcache_because(key.toString()).warn();
}
} else {
cache.remove(key);
}
}
}
}
}
return old;
}
private void checkHistoryParsing(
final int numMaps, final int numReduces, final int numSuccessfulMaps) throws Exception {
Configuration conf = new Configuration();
conf.set(MRJobConfig.USER_NAME, System.getProperty("user.name"));
long amStartTimeEst = System.currentTimeMillis();
conf.setClass(
CommonConfigurationKeysPublic.NET_TOPOLOGY_NODE_SWITCH_MAPPING_IMPL_KEY,
MyResolver.class,
DNSToSwitchMapping.class);
RackResolver.init(conf);
MRApp app = new MRAppWithHistory(numMaps, numReduces, true, this.getClass().getName(), true);
app.submit(conf);
Job job = app.getContext().getAllJobs().values().iterator().next();
JobId jobId = job.getID();
LOG.info("JOBID is " + TypeConverter.fromYarn(jobId).toString());
app.waitForState(job, JobState.SUCCEEDED);
// make sure all events are flushed
app.waitForState(Service.STATE.STOPPED);
String jobhistoryDir = JobHistoryUtils.getHistoryIntermediateDoneDirForUser(conf);
FileContext fc = null;
try {
fc = FileContext.getFileContext(conf);
} catch (IOException ioe) {
LOG.info("Can not get FileContext", ioe);
throw (new Exception("Can not get File Context"));
}
if (numMaps == numSuccessfulMaps) {
String summaryFileName = JobHistoryUtils.getIntermediateSummaryFileName(jobId);
Path summaryFile = new Path(jobhistoryDir, summaryFileName);
String jobSummaryString = getJobSummary(fc, summaryFile);
Assert.assertNotNull(jobSummaryString);
Assert.assertTrue(jobSummaryString.contains("resourcesPerMap=100"));
Assert.assertTrue(jobSummaryString.contains("resourcesPerReduce=100"));
Map<String, String> jobSummaryElements = new HashMap<String, String>();
StringTokenizer strToken = new StringTokenizer(jobSummaryString, ",");
while (strToken.hasMoreTokens()) {
String keypair = strToken.nextToken();
jobSummaryElements.put(keypair.split("=")[0], keypair.split("=")[1]);
}
Assert.assertEquals(
"JobId does not match", jobId.toString(), jobSummaryElements.get("jobId"));
Assert.assertEquals("JobName does not match", "test", jobSummaryElements.get("jobName"));
Assert.assertTrue(
"submitTime should not be 0", Long.parseLong(jobSummaryElements.get("submitTime")) != 0);
Assert.assertTrue(
"launchTime should not be 0", Long.parseLong(jobSummaryElements.get("launchTime")) != 0);
Assert.assertTrue(
"firstMapTaskLaunchTime should not be 0",
Long.parseLong(jobSummaryElements.get("firstMapTaskLaunchTime")) != 0);
Assert.assertTrue(
"firstReduceTaskLaunchTime should not be 0",
Long.parseLong(jobSummaryElements.get("firstReduceTaskLaunchTime")) != 0);
Assert.assertTrue(
"finishTime should not be 0", Long.parseLong(jobSummaryElements.get("finishTime")) != 0);
Assert.assertEquals(
"Mismatch in num map slots",
numSuccessfulMaps,
Integer.parseInt(jobSummaryElements.get("numMaps")));
Assert.assertEquals(
"Mismatch in num reduce slots",
numReduces,
Integer.parseInt(jobSummaryElements.get("numReduces")));
Assert.assertEquals(
"User does not match", System.getProperty("user.name"), jobSummaryElements.get("user"));
Assert.assertEquals("Queue does not match", "default", jobSummaryElements.get("queue"));
Assert.assertEquals("Status does not match", "SUCCEEDED", jobSummaryElements.get("status"));
}
JobHistory jobHistory = new JobHistory();
jobHistory.init(conf);
HistoryFileInfo fileInfo = jobHistory.getJobFileInfo(jobId);
JobInfo jobInfo;
long numFinishedMaps;
synchronized (fileInfo) {
Path historyFilePath = fileInfo.getHistoryFile();
FSDataInputStream in = null;
LOG.info("JobHistoryFile is: " + historyFilePath);
try {
in = fc.open(fc.makeQualified(historyFilePath));
} catch (IOException ioe) {
LOG.info("Can not open history file: " + historyFilePath, ioe);
throw (new Exception("Can not open History File"));
}
JobHistoryParser parser = new JobHistoryParser(in);
final EventReader realReader = new EventReader(in);
EventReader reader = Mockito.mock(EventReader.class);
if (numMaps == numSuccessfulMaps) {
reader = realReader;
} else {
final AtomicInteger numFinishedEvents = new AtomicInteger(0); // Hack!
Mockito.when(reader.getNextEvent())
.thenAnswer(
new Answer<HistoryEvent>() {
public HistoryEvent answer(InvocationOnMock invocation) throws IOException {
HistoryEvent event = realReader.getNextEvent();
if (event instanceof TaskFinishedEvent) {
numFinishedEvents.incrementAndGet();
}
if (numFinishedEvents.get() <= numSuccessfulMaps) {
return event;
} else {
throw new IOException("test");
}
}
});
}
jobInfo = parser.parse(reader);
numFinishedMaps = computeFinishedMaps(jobInfo, numMaps, numSuccessfulMaps);
if (numFinishedMaps != numMaps) {
Exception parseException = parser.getParseException();
Assert.assertNotNull("Didn't get expected parse exception", parseException);
}
}
Assert.assertEquals(
"Incorrect username ", System.getProperty("user.name"), jobInfo.getUsername());
Assert.assertEquals("Incorrect jobName ", "test", jobInfo.getJobname());
Assert.assertEquals("Incorrect queuename ", "default", jobInfo.getJobQueueName());
Assert.assertEquals("incorrect conf path", "test", jobInfo.getJobConfPath());
Assert.assertEquals("incorrect finishedMap ", numSuccessfulMaps, numFinishedMaps);
Assert.assertEquals("incorrect finishedReduces ", numReduces, jobInfo.getFinishedReduces());
Assert.assertEquals("incorrect uberized ", job.isUber(), jobInfo.getUberized());
Map<TaskID, TaskInfo> allTasks = jobInfo.getAllTasks();
int totalTasks = allTasks.size();
Assert.assertEquals("total number of tasks is incorrect ", (numMaps + numReduces), totalTasks);
// Verify aminfo
Assert.assertEquals(1, jobInfo.getAMInfos().size());
Assert.assertEquals(MRApp.NM_HOST, jobInfo.getAMInfos().get(0).getNodeManagerHost());
AMInfo amInfo = jobInfo.getAMInfos().get(0);
Assert.assertEquals(MRApp.NM_PORT, amInfo.getNodeManagerPort());
Assert.assertEquals(MRApp.NM_HTTP_PORT, amInfo.getNodeManagerHttpPort());
Assert.assertEquals(1, amInfo.getAppAttemptId().getAttemptId());
Assert.assertEquals(
amInfo.getAppAttemptId(), amInfo.getContainerId().getApplicationAttemptId());
Assert.assertTrue(
amInfo.getStartTime() <= System.currentTimeMillis()
&& amInfo.getStartTime() >= amStartTimeEst);
ContainerId fakeCid = BuilderUtils.newContainerId(-1, -1, -1, -1);
// Assert at taskAttempt level
for (TaskInfo taskInfo : allTasks.values()) {
int taskAttemptCount = taskInfo.getAllTaskAttempts().size();
Assert.assertEquals("total number of task attempts ", 1, taskAttemptCount);
TaskAttemptInfo taInfo = taskInfo.getAllTaskAttempts().values().iterator().next();
Assert.assertNotNull(taInfo.getContainerId());
// Verify the wrong ctor is not being used. Remove after mrv1 is removed.
Assert.assertFalse(taInfo.getContainerId().equals(fakeCid));
}
// Deep compare Job and JobInfo
for (Task task : job.getTasks().values()) {
TaskInfo taskInfo = allTasks.get(TypeConverter.fromYarn(task.getID()));
Assert.assertNotNull("TaskInfo not found", taskInfo);
for (TaskAttempt taskAttempt : task.getAttempts().values()) {
TaskAttemptInfo taskAttemptInfo =
taskInfo.getAllTaskAttempts().get(TypeConverter.fromYarn((taskAttempt.getID())));
Assert.assertNotNull("TaskAttemptInfo not found", taskAttemptInfo);
Assert.assertEquals(
"Incorrect shuffle port for task attempt",
taskAttempt.getShufflePort(),
taskAttemptInfo.getShufflePort());
if (numMaps == numSuccessfulMaps) {
Assert.assertEquals(MRApp.NM_HOST, taskAttemptInfo.getHostname());
Assert.assertEquals(MRApp.NM_PORT, taskAttemptInfo.getPort());
// Verify rack-name
Assert.assertEquals("rack-name is incorrect", taskAttemptInfo.getRackname(), RACK_NAME);
}
}
}
}
private synchronized void moveToDone() throws IOException {
if (LogGlobal.isDebugEnabled()) {
/* LOG.debug("moveToDone: "+historyFile) */
LOG.movetodone(historyFile.toString()).debug();
}
if (!isMovePending()) {
// It was either deleted or is already in done. Either way do nothing
if (LogGlobal.isDebugEnabled()) {
/* LOG.debug("Move no longer pending") */
LOG.move_longer_pending().debug();
}
return;
}
try {
long completeTime = jobIndexInfo.getFinishTime();
if (completeTime == 0) {
completeTime = System.currentTimeMillis();
}
JobId jobId = jobIndexInfo.getJobId();
List<Path> paths = new ArrayList<Path>(2);
if (historyFile == null) {
/* LOG.info("No file for job-history with "+jobId+" found in cache!") */
LOG.file_for_job_history_with_found(jobId.toString()).info();
} else {
paths.add(historyFile);
}
if (confFile == null) {
/* LOG.info("No file for jobConf with "+jobId+" found in cache!") */
LOG.file_for_jobconf_with_found_cache(jobId.toString()).info();
} else {
paths.add(confFile);
}
if (summaryFile == null) {
/* LOG.info("No summary file for job: "+jobId) */
LOG.summary_file_for_job(jobId.toString()).info();
} else {
String jobSummaryString = getJobSummary(intermediateDoneDirFc, summaryFile);
LOG.summary_file_for_job(jobSummaryString);
/* LOG.info("Deleting JobSummary file: ["+summaryFile+"]") */
LOG.deleting_jobsummary_file(summaryFile.toString()).info();
intermediateDoneDirFc.delete(summaryFile, false);
summaryFile = null;
}
Path targetDir = canonicalHistoryLogPath(jobId, completeTime);
addDirectoryToSerialNumberIndex(targetDir);
makeDoneSubdir(targetDir);
if (historyFile != null) {
Path toPath = doneDirFc.makeQualified(new Path(targetDir, historyFile.getName()));
if (!toPath.equals(historyFile)) {
moveToDoneNow(historyFile, toPath);
historyFile = toPath;
}
}
if (confFile != null) {
Path toPath = doneDirFc.makeQualified(new Path(targetDir, confFile.getName()));
if (!toPath.equals(confFile)) {
moveToDoneNow(confFile, toPath);
confFile = toPath;
}
}
state = HistoryInfoState.IN_DONE;
} catch (Throwable t) {
/* LOG.error("Error while trying to move a job to done",t) */
LOG.error_while_trying_move_job_done(t.toString()).error();
this.state = HistoryInfoState.MOVE_FAILED;
}
}