@Override
HeartbeatResponse transmitHeartBeat(long now) throws IOException {
HeartbeatResponse response = super.transmitHeartBeat(now);
LOG.info("WaitingTaskTracker waiting");
// wait forever
UtilsForTests.waitFor(Long.MAX_VALUE);
throw new IOException("WaitingTaskTracker interrupted. Bailing out");
}
// wait till the job retires
private void waitTillRetire(JobID id, JobTracker jobtracker) {
// wait for job to get retired
JobInProgress job = jobtracker.getJob(id);
for (int i = 0; i < 10 && job != null; i++) {
UtilsForTests.waitFor(1000);
job = jobtracker.getJob(id);
}
assertNull("Job did not retire", job);
}
/**
* Set the invalid mapred local directory location and run the job. Verify the job status.
*
* @throws Exception - if an error occurs.
*/
@Test
public void testJobStatusForInvalidTaskControllerConf() throws Exception {
conf = remoteJTClient.getDaemonConf();
if (conf.get("mapred.task.tracker.task-controller")
.equals("org.apache.hadoop.mapred.LinuxTaskController")) {
StringBuffer mapredLocalDir = new StringBuffer();
LOG.info("JobConf.MAPRED_LOCAL_DIR_PROPERTY:" + conf.get(JobConf.MAPRED_LOCAL_DIR_PROPERTY));
mapredLocalDir.append(conf.get(JobConf.MAPRED_LOCAL_DIR_PROPERTY));
mapredLocalDir.append(",");
mapredLocalDir.append("/mapred/local");
String jobArgs[] = {
"-D",
"mapred.local.dir=" + mapredLocalDir.toString(),
"-m",
"1",
"-r",
"1",
"-mt",
"1000",
"-rt",
"1000",
"-recordt",
"100"
};
SleepJob job = new SleepJob();
JobConf jobConf = new JobConf(conf);
int exitStatus = ToolRunner.run(jobConf, job, jobArgs);
Assert.assertEquals("Exit Code:", 0, exitStatus);
UtilsForTests.waitFor(100);
JobClient jobClient = jtClient.getClient();
JobID jobId = jobClient.getAllJobs()[0].getJobID();
LOG.info("JobId:" + jobId);
if (jobId != null) {
JobInfo jInfo = remoteJTClient.getJobInfo(jobId);
Assert.assertEquals(
"Job has not been succeeded", jInfo.getStatus().getRunState(), JobStatus.SUCCEEDED);
}
} else {
Assert.assertTrue("Linux Task controller not found.", false);
}
}
@SuppressWarnings("unchecked")
public void testCommitter() throws Exception {
JobConf job = new JobConf();
setConfForFileOutputCommitter(job);
JobContext jContext = new JobContextImpl(job, taskID.getJobID());
TaskAttemptContext tContext = new TaskAttemptContextImpl(job, taskID);
FileOutputCommitter committer = new FileOutputCommitter();
FileOutputFormat.setWorkOutputPath(job, committer.getTempTaskOutputPath(tContext));
committer.setupJob(jContext);
committer.setupTask(tContext);
String file = "test.txt";
// A reporter that does nothing
Reporter reporter = Reporter.NULL;
// write output
FileSystem localFs = FileSystem.getLocal(job);
TextOutputFormat theOutputFormat = new TextOutputFormat();
RecordWriter theRecordWriter = theOutputFormat.getRecordWriter(localFs, job, file, reporter);
writeOutput(theRecordWriter, reporter);
// do commit
committer.commitTask(tContext);
committer.commitJob(jContext);
// validate output
File expectedFile = new File(new Path(outDir, file).toString());
StringBuffer expectedOutput = new StringBuffer();
expectedOutput.append(key1).append('\t').append(val1).append("\n");
expectedOutput.append(val1).append("\n");
expectedOutput.append(val2).append("\n");
expectedOutput.append(key2).append("\n");
expectedOutput.append(key1).append("\n");
expectedOutput.append(key2).append('\t').append(val2).append("\n");
String output = UtilsForTests.slurp(expectedFile);
assertEquals(output, expectedOutput.toString());
FileUtil.fullyDelete(new File(outDir.toString()));
}
private JobID validateJobRetire(JobConf jobConf, Path inDir, Path outDir, JobTracker jobtracker)
throws IOException {
RunningJob rj = UtilsForTests.runJob(jobConf, inDir, outDir, 0, 0);
rj.waitForCompletion();
assertTrue(rj.isSuccessful());
JobID id = rj.getID();
// wait for job to get retired
waitTillRetire(id, jobtracker);
RetireJobInfo retired = jobtracker.retireJobs.get(id);
assertTrue(
"History url not set",
retired.getHistoryFile() != null && retired.getHistoryFile().length() > 0);
assertNotNull("Job is not in cache", jobtracker.getJobStatus(id));
// get the job conf filename
String name = jobtracker.getLocalJobFilePath(id);
File file = new File(name);
assertFalse("JobConf file not deleted", file.exists());
// test redirection
URL jobUrl = new URL(rj.getTrackingURL());
HttpURLConnection conn = (HttpURLConnection) jobUrl.openConnection();
conn.setInstanceFollowRedirects(false);
conn.connect();
assertEquals(HttpURLConnection.HTTP_MOVED_TEMP, conn.getResponseCode());
conn.disconnect();
URL redirectedUrl = new URL(conn.getHeaderField("Location"));
conn = (HttpURLConnection) redirectedUrl.openConnection();
conn.connect();
assertEquals(HttpURLConnection.HTTP_OK, conn.getResponseCode());
conn.disconnect();
return id;
}
void testNormalWriteAndRead(
final int segments, final int length, final int byteMax, final CompressionCodec codec)
throws Exception {
Random random = new Random(37);
// Prepare data for testing
SimpleSeekableFormat.Buffer[] data = new SimpleSeekableFormat.Buffer[segments];
for (int s = 0; s < segments; s++) {
data[s] = createBuffer(random, length, byteMax);
}
// Write data to dataSegmentOutput
ByteArrayOutputStream dataSegmentOutput = new ByteArrayOutputStream();
for (int s = 0; s < segments; s++) {
DataSegmentWriter writer = new DataSegmentWriter(data[s], codec, null);
writer.writeTo(new DataOutputStream(dataSegmentOutput));
}
// Read dataSegments back to a Buffer
byte[][] newData = new byte[segments][];
ByteArrayInputStream dataSegmentInput =
new ByteArrayInputStream(dataSegmentOutput.toByteArray());
Configuration conf = new Configuration();
HashMap<Text, Decompressor> decompressorCache = new HashMap<Text, Decompressor>();
for (int s = 0; s < segments; s++) {
DataSegmentReader reader =
new DataSegmentReader(new DataInputStream(dataSegmentInput), conf, decompressorCache);
newData[s] = IOUtils.toByteArray(reader.getInputStream());
}
// Compare data and newData
for (int s = 0; s < segments; s++) {
UtilsForTests.assertArrayEquals("Segment[" + s + "]", data[s].toByteArray(), newData[s]);
}
}
/** Check refreshNodes for decommissioning blacklisted nodes. */
public void testBlacklistedNodeDecommissioning() throws Exception {
LOG.info("Testing blacklisted node decommissioning");
MiniMRCluster mr = null;
JobTracker jt = null;
try {
// start mini mr
JobConf jtConf = new JobConf();
jtConf.set("mapred.max.tracker.blacklists", "1");
mr = new MiniMRCluster(0, 0, 2, "file:///", 1, null, null, null, jtConf);
jt = mr.getJobTrackerRunner().getJobTracker();
assertEquals("Trackers not up", 2, jt.taskTrackers().size());
// validate the total tracker count
assertEquals(
"Active tracker count mismatch", 2, jt.getClusterStatus(false).getTaskTrackers());
// validate blacklisted count
assertEquals(
"Blacklisted tracker count mismatch",
0,
jt.getClusterStatus(false).getBlacklistedTrackers());
// run a failing job to blacklist the tracker
JobConf jConf = mr.createJobConf();
jConf.set("mapred.max.tracker.failures", "1");
jConf.setJobName("test-job-fail-once");
jConf.setMapperClass(FailOnceMapper.class);
jConf.setReducerClass(IdentityReducer.class);
jConf.setNumMapTasks(1);
jConf.setNumReduceTasks(0);
RunningJob job =
UtilsForTests.runJob(jConf, new Path(TEST_DIR, "in"), new Path(TEST_DIR, "out"));
job.waitForCompletion();
// validate the total tracker count
assertEquals(
"Active tracker count mismatch", 1, jt.getClusterStatus(false).getTaskTrackers());
// validate blacklisted count
assertEquals(
"Blacklisted tracker count mismatch",
1,
jt.getClusterStatus(false).getBlacklistedTrackers());
// find the blacklisted tracker
String trackerName = null;
for (TaskTrackerStatus status : jt.taskTrackers()) {
if (jt.isBlacklisted(status.getTrackerName())) {
trackerName = status.getTrackerName();
break;
}
}
// get the hostname
String hostToDecommission = JobInProgress.convertTrackerNameToHostName(trackerName);
LOG.info("Decommissioning tracker " + hostToDecommission);
// decommission the node
HashSet<String> decom = new HashSet<String>(1);
decom.add(hostToDecommission);
jt.decommissionNodes(decom);
// validate
// check the cluster status and tracker size
assertEquals(
"Tracker is not lost upon host decommissioning",
1,
jt.getClusterStatus(false).getTaskTrackers());
assertEquals(
"Blacklisted tracker count incorrect in cluster status " + "after decommissioning",
0,
jt.getClusterStatus(false).getBlacklistedTrackers());
assertEquals("Tracker is not lost upon host decommissioning", 1, jt.taskTrackers().size());
} finally {
if (mr != null) {
mr.shutdown();
mr = null;
jt = null;
FileUtil.fullyDelete(new File(TEST_DIR.toString()));
}
}
}
/**
* Test job retire with tasks that report their *first* status only after the job retires. Steps :
* - Start a mini-mr cluster with 1 task-tracker having only map slots. Note that this
* task-tracker will take care of setup/cleanup and the map tasks. - Submit a job with 1 map task
* and 1 reduce task - Wait for the job to finish the map task - Start a 2nd tracker that waits
* for a long time after contacting the JT. - Wait for the 2nd tracker to get stuck - Kill the job
* - Wait for the job to retire - Check if the tip mappings are cleaned up.
*/
public void testJobRetireWithUnreportedTasks() throws Exception {
MiniMRCluster mr = null;
try {
JobConf conf = new JobConf();
// set the num-map-slots to 1 so that no reduce tasks but setup/cleanup
// can run on it
conf.setInt("mapred.tasktracker.map.tasks.maximum", 1);
conf.setInt("mapred.tasktracker.reduce.tasks.maximum", 0);
mr = startCluster(conf, 1);
JobTracker jobtracker = mr.getJobTrackerRunner().getJobTracker();
RunningJob job =
UtilsForTests.runJob(
mr.createJobConf(), new Path(testDir, "in-1"), new Path(testDir, "out-1"), 1, 1);
JobID id = JobID.downgrade(job.getID());
JobInProgress jip = jobtracker.getJob(id);
// wait 100 secs for the job to complete its map task
for (int i = 0; i < 1000 && jip.finishedMaps() < 1; i++) {
UtilsForTests.waitFor(100);
}
assertEquals(jip.finishedMaps(), 1);
// start a tracker that will wait
LOG.info("Adding a waiting tracker");
TaskTrackerRunner testTrackerRunner =
mr.new TaskTrackerRunner(1, 1, null, mr.createJobConf()) {
@Override
TaskTracker createTaskTracker(JobConf conf) throws InterruptedException, IOException {
return new WaitingTaskTracker(conf);
}
};
mr.addTaskTracker(testTrackerRunner);
LOG.info("Waiting tracker added");
WaitingTaskTracker testTT = (WaitingTaskTracker) testTrackerRunner.getTaskTracker();
// wait 100 secs for the newly started task-tracker to join
for (int i = 0; i < 1000 && jobtracker.taskTrackers().size() < 2; i++) {
UtilsForTests.waitFor(100);
}
assertEquals(jobtracker.taskTrackers().size(), 2);
LOG.info("Cluster is now up with 2 trackers");
// stop the test-tt as its no longer required
mr.stopTaskTracker(mr.getTaskTrackerID(testTT.getName()));
// 1 reduce task should be scheduled
assertEquals("TestTT contacted but no reduce task scheduled on it", 1, jip.runningReduces());
// kill the job
LOG.info("Killing job " + id);
job.killJob();
// check if the reduce task attempt status is missing
TaskInProgress tip = jip.getTasks(TaskType.REDUCE)[0];
assertNull(tip.getTaskStatus(tip.getAllTaskAttemptIDs()[0]));
// wait for the job to retire
waitTillRetire(id, jobtracker);
// check the taskidToTIPMap
for (TaskAttemptID tid : jobtracker.taskidToTIPMap.keySet()) {
LOG.info("TaskidToTIP : " + tid);
}
assertEquals("'taskid' to TIP mapping still exists", 0, jobtracker.taskidToTIPMap.size());
} finally {
if (mr != null) {
mr.shutdown();
}
// cleanup
FileUtil.fullyDelete(new File(testDir.toString()));
}
}
@Test
public void mrRun() throws Exception {
FileSystem fs = dfsCluster.getFileSystem();
Path inDir = fs.makeQualified(new Path("/user/testing/testMapperReducer/input"));
fs.delete(inDir, true);
String DATADIR = "/user/testing/testMapperReducer/data";
Path dataDir = fs.makeQualified(new Path(DATADIR));
fs.delete(dataDir, true);
Path outDir = fs.makeQualified(new Path("/user/testing/testMapperReducer/output"));
fs.delete(outDir, true);
assertTrue(fs.mkdirs(inDir));
Path INPATH = new Path(inDir, "input.txt");
OutputStream os = fs.create(INPATH);
Writer wr = new OutputStreamWriter(os, StandardCharsets.UTF_8);
wr.write(DATADIR + "/" + inputAvroFile);
wr.close();
assertTrue(fs.mkdirs(dataDir));
fs.copyFromLocalFile(new Path(DOCUMENTS_DIR, inputAvroFile), dataDir);
JobConf jobConf = getJobConf();
jobConf.set("jobclient.output.filter", "ALL");
if (ENABLE_LOCAL_JOB_RUNNER) { // enable Hadoop LocalJobRunner; this enables to run in debugger
// and set breakpoints
jobConf.set("mapred.job.tracker", "local");
}
jobConf.setMaxMapAttempts(1);
jobConf.setMaxReduceAttempts(1);
jobConf.setJar(SEARCH_ARCHIVES_JAR);
int shards = 2;
int maxReducers = Integer.MAX_VALUE;
if (ENABLE_LOCAL_JOB_RUNNER) {
// local job runner has a couple of limitations: only one reducer is supported and the
// DistributedCache doesn't work.
// see http://blog.cloudera.com/blog/2009/07/advice-on-qa-testing-your-mapreduce-jobs/
maxReducers = 1;
shards = 1;
}
String[] args =
new String[] {
"--morphline-file=" + tempDir + "/test-morphlines/solrCellDocumentTypes.conf",
"--morphline-id=morphline1",
"--solr-home-dir=" + MINIMR_CONF_DIR.getAbsolutePath(),
"--output-dir=" + outDir.toString(),
"--shards=" + shards,
"--verbose",
numRuns % 2 == 0 ? "--input-list=" + INPATH.toString() : dataDir.toString(),
numRuns % 3 == 0
? "--reducers=" + shards
: (numRuns % 3 == 1 ? "--reducers=-1" : "--reducers=" + Math.min(8, maxReducers))
};
if (numRuns % 3 == 2) {
args = concat(args, new String[] {"--fanout=2"});
}
if (numRuns == 0) {
// force (slow) MapReduce based randomization to get coverage for that as well
args =
concat(
new String[] {"-D", MapReduceIndexerTool.MAIN_MEMORY_RANDOMIZATION_THRESHOLD + "=-1"},
args);
}
MapReduceIndexerTool tool = createTool();
int res = ToolRunner.run(jobConf, tool, args);
assertEquals(0, res);
Job job = tool.job;
assertTrue(job.isComplete());
assertTrue(job.isSuccessful());
if (numRuns % 3 != 2) {
// Only run this check if mtree merge is disabled.
// With mtree merge enabled the BatchWriter counters aren't available anymore because
// variable "job" now refers to the merge job rather than the indexing job
assertEquals(
"Invalid counter "
+ SolrRecordWriter.class.getName()
+ "."
+ SolrCounters.DOCUMENTS_WRITTEN,
count,
job.getCounters()
.findCounter(SolrCounters.class.getName(), SolrCounters.DOCUMENTS_WRITTEN.toString())
.getValue());
}
// Check the output is as expected
outDir = new Path(outDir, MapReduceIndexerTool.RESULTS_DIR);
Path[] outputFiles = FileUtil.stat2Paths(fs.listStatus(outDir));
System.out.println("outputfiles:" + Arrays.toString(outputFiles));
UtilsForTests.validateSolrServerDocumentCount(MINIMR_CONF_DIR, fs, outDir, count, shards);
// run again with --dryrun mode:
tool = createTool();
args = concat(args, new String[] {"--dry-run"});
res = ToolRunner.run(jobConf, tool, args);
assertEquals(0, res);
numRuns++;
}
