@SuppressWarnings("unchecked")
@Override
public int run(String[] args) throws Exception {
MyOptions options = new MyOptions(args, getConf());
traceWriter = options.clazzTraceOutputter.newInstance();
traceWriter.init(options.traceOutput, getConf());
topologyWriter = new DefaultOutputter<LoggedNetworkTopology>();
topologyWriter.init(options.topologyOutput, getConf());
try {
JobBuilder jobBuilder = null;
for (Path p : options.inputs) {
InputDemuxer inputDemuxer = options.inputDemuxerClass.newInstance();
try {
inputDemuxer.bindTo(p, getConf());
} catch (IOException e) {
LOG.warn("Unable to bind Path " + p + " . Skipping...", e);
continue;
}
Pair<String, InputStream> filePair = null;
try {
while ((filePair = inputDemuxer.getNext()) != null) {
RewindableInputStream ris = new RewindableInputStream(filePair.second());
JobHistoryParser parser = null;
try {
String jobID = JobHistoryUtils.extractJobID(filePair.first());
if (jobID == null) {
LOG.warn("File skipped: Invalid file name: " + filePair.first());
continue;
}
if ((jobBuilder == null) || (!jobBuilder.getJobID().equals(jobID))) {
if (jobBuilder != null) {
traceWriter.output(jobBuilder.build());
}
jobBuilder = new JobBuilder(jobID);
}
if (JobHistoryUtils.isJobConfXml(filePair.first())) {
processJobConf(JobConfigurationParser.parse(ris.rewind()), jobBuilder);
} else {
parser = JobHistoryParserFactory.getParser(ris);
if (parser == null) {
LOG.warn("File skipped: Cannot find suitable parser: " + filePair.first());
} else {
processJobHistory(parser, jobBuilder);
}
}
} finally {
if (parser == null) {
ris.close();
} else {
parser.close();
parser = null;
}
}
}
} catch (Throwable t) {
if (filePair != null) {
LOG.warn(
"TraceBuilder got an error while processing the [possibly virtual] file "
+ filePair.first()
+ " within Path "
+ p,
t);
}
} finally {
inputDemuxer.close();
}
}
if (jobBuilder != null) {
traceWriter.output(jobBuilder.build());
jobBuilder = null;
} else {
LOG.warn("No job found in traces: ");
}
topologyWriter.output(topologyBuilder.build());
} finally {
traceWriter.close();
topologyWriter.close();
}
return 0;
}
/** Test if {@link CurrentJHParser} can read events from current JH files. */
@Test
public void testCurrentJHParser() throws Exception {
final Configuration conf = new Configuration();
final FileSystem lfs = FileSystem.getLocal(conf);
final Path rootTempDir =
new Path(System.getProperty("test.build.data", "/tmp")).makeQualified(lfs);
final Path tempDir = new Path(rootTempDir, "TestCurrentJHParser");
lfs.delete(tempDir, true);
String queueName = "testQueue";
// Run a MR job
// create a MR cluster
conf.setInt("mapred.tasktracker.map.tasks.maximum", 1);
conf.setInt("mapred.tasktracker.reduce.tasks.maximum", 1);
conf.set("mapred.queue.names", queueName);
MiniMRCluster mrCluster = new MiniMRCluster(1, "file:///", 1, null, null, new JobConf(conf));
// run a job
Path inDir = new Path(tempDir, "input");
Path outDir = new Path(tempDir, "output");
JobHistoryParser parser = null;
RewindableInputStream ris = null;
ArrayList<String> seenEvents = new ArrayList<String>(10);
RunningJob rJob = null;
try {
JobConf jobConf = mrCluster.createJobConf();
jobConf.setQueueName(queueName);
// construct a job with 1 map and 1 reduce task.
rJob = UtilsForTests.runJob(jobConf, inDir, outDir, 1, 1);
rJob.waitForCompletion();
assertTrue("Job failed", rJob.isSuccessful());
JobID id = rJob.getID();
// get the jobhistory filepath
Path inputPath =
new Path(JobHistory.getHistoryFilePath(org.apache.hadoop.mapred.JobID.downgrade(id)));
// wait for 10 secs for the jobhistory file to move into the done folder
for (int i = 0; i < 100; ++i) {
if (lfs.exists(inputPath)) {
break;
}
TimeUnit.MILLISECONDS.wait(100);
}
assertTrue("Missing job history file", lfs.exists(inputPath));
InputDemuxer inputDemuxer = new DefaultInputDemuxer();
inputDemuxer.bindTo(inputPath, conf);
Pair<String, InputStream> filePair = inputDemuxer.getNext();
assertNotNull(filePair);
ris = new RewindableInputStream(filePair.second());
// Test if the JobHistoryParserFactory can detect the parser correctly
parser = JobHistoryParserFactory.getParser(ris);
// Get ParsedJob
String jobId = TraceBuilder.extractJobID(filePair.first());
JobBuilder builder = new JobBuilder(jobId);
HistoryEvent e;
while ((e = parser.nextEvent()) != null) {
String eventString = e.getEventType().toString();
System.out.println(eventString);
seenEvents.add(eventString);
if (builder != null) {
builder.process(e);
}
}
ParsedJob parsedJob = builder.build();
// validate the obtainXXX api of ParsedJob, ParsedTask and
// ParsedTaskAttempt.
validateParsedJob(parsedJob, 1, 1, queueName);
} finally {
// stop the MR cluster
mrCluster.shutdown();
if (ris != null) {
ris.close();
}
if (parser != null) {
parser.close();
}
// cleanup the filesystem
lfs.delete(tempDir, true);
}
// Check against the gold standard
System.out.println("testCurrentJHParser validating using gold std ");
String[] goldLines =
new String[] {
"JOB_SUBMITTED",
"JOB_PRIORITY_CHANGED",
"JOB_STATUS_CHANGED",
"JOB_INITED",
"JOB_INFO_CHANGED",
"TASK_STARTED",
"MAP_ATTEMPT_STARTED",
"MAP_ATTEMPT_FINISHED",
"MAP_ATTEMPT_FINISHED",
"TASK_UPDATED",
"TASK_FINISHED",
"JOB_STATUS_CHANGED",
"TASK_STARTED",
"MAP_ATTEMPT_STARTED",
"MAP_ATTEMPT_FINISHED",
"MAP_ATTEMPT_FINISHED",
"TASK_UPDATED",
"TASK_FINISHED",
"TASK_STARTED",
"MAP_ATTEMPT_STARTED",
"MAP_ATTEMPT_FINISHED",
"REDUCE_ATTEMPT_FINISHED",
"TASK_UPDATED",
"TASK_FINISHED",
"TASK_STARTED",
"MAP_ATTEMPT_STARTED",
"MAP_ATTEMPT_FINISHED",
"MAP_ATTEMPT_FINISHED",
"TASK_UPDATED",
"TASK_FINISHED",
"JOB_STATUS_CHANGED",
"JOB_FINISHED"
};
// Check the output with gold std
assertEquals("Size mismatch", goldLines.length, seenEvents.size());
int index = 0;
for (String goldLine : goldLines) {
assertEquals("Content mismatch", goldLine, seenEvents.get(index++));
}
}