use of org.apache.tez.dag.api.client.StatusGetOpts in project tez by apache.
the class TestOrderedWordCount method run.
@Override
public int run(String[] args) throws Exception {
Configuration conf = getConf();
String[] otherArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
boolean generateSplitsInClient;
SplitsInClientOptionParser splitCmdLineParser = new SplitsInClientOptionParser();
try {
generateSplitsInClient = splitCmdLineParser.parse(otherArgs, false);
otherArgs = splitCmdLineParser.getRemainingArgs();
} catch (ParseException e1) {
System.err.println("Invalid options");
printUsage();
return 2;
}
boolean useTezSession = conf.getBoolean("USE_TEZ_SESSION", true);
long interJobSleepTimeout = conf.getInt("INTER_JOB_SLEEP_INTERVAL", 0) * 1000;
boolean retainStagingDir = conf.getBoolean("RETAIN_STAGING_DIR", false);
boolean useMRSettings = conf.getBoolean("USE_MR_CONFIGS", true);
// TODO needs to use auto reduce parallelism
int intermediateNumReduceTasks = conf.getInt("IREDUCE_NUM_TASKS", 2);
int maxDataLengthThroughIPC = conf.getInt(MAX_IPC_DATA_LENGTH, -1);
int exceedDataLimit = conf.getInt(EXCEED_IPC_DATA_LIMIT, 3);
if (maxDataLengthThroughIPC > 0) {
conf.setInt(CommonConfigurationKeys.IPC_MAXIMUM_DATA_LENGTH, maxDataLengthThroughIPC * 1024 * 1024);
}
if (((otherArgs.length % 2) != 0) || (!useTezSession && otherArgs.length != 2)) {
printUsage();
return 2;
}
List<String> inputPaths = new ArrayList<String>();
List<String> outputPaths = new ArrayList<String>();
TezConfiguration tezConf = new TezConfiguration(conf);
for (int i = 0; i < otherArgs.length; i += 2) {
FileSystem inputPathFs = new Path(otherArgs[i]).getFileSystem(tezConf);
inputPaths.add(inputPathFs.makeQualified(new Path(otherArgs[i])).toString());
FileSystem outputPathFs = new Path(otherArgs[i + 1]).getFileSystem(tezConf);
outputPaths.add(outputPathFs.makeQualified(new Path(otherArgs[i + 1])).toString());
}
UserGroupInformation.setConfiguration(conf);
HadoopShim hadoopShim = new HadoopShimsLoader(tezConf).getHadoopShim();
TestOrderedWordCount instance = new TestOrderedWordCount();
FileSystem fs = FileSystem.get(conf);
String stagingDirStr = conf.get(TezConfiguration.TEZ_AM_STAGING_DIR, TezConfiguration.TEZ_AM_STAGING_DIR_DEFAULT) + Path.SEPARATOR + Long.toString(System.currentTimeMillis());
Path stagingDir = new Path(stagingDirStr);
FileSystem pathFs = stagingDir.getFileSystem(tezConf);
pathFs.mkdirs(new Path(stagingDirStr));
tezConf.set(TezConfiguration.TEZ_AM_STAGING_DIR, stagingDirStr);
stagingDir = pathFs.makeQualified(new Path(stagingDirStr));
TokenCache.obtainTokensForNamenodes(instance.credentials, new Path[] { stagingDir }, conf);
TezClientUtils.ensureStagingDirExists(tezConf, stagingDir);
if (useTezSession) {
LOG.info("Creating Tez Session");
tezConf.setBoolean(TezConfiguration.TEZ_AM_SESSION_MODE, true);
} else {
tezConf.setBoolean(TezConfiguration.TEZ_AM_SESSION_MODE, false);
}
TezClient tezSession = TezClient.create("OrderedWordCountSession", tezConf, null, instance.credentials);
tezSession.start();
if (tezSession.getAppMasterApplicationId() != null) {
TezUtilsInternal.setHadoopCallerContext(hadoopShim, tezSession.getAppMasterApplicationId());
}
DAGStatus dagStatus = null;
DAGClient dagClient = null;
String[] vNames = { "initialmap", "intermediate_reducer", "finalreduce" };
Set<StatusGetOpts> statusGetOpts = EnumSet.of(StatusGetOpts.GET_COUNTERS);
try {
for (int dagIndex = 1; dagIndex <= inputPaths.size(); ++dagIndex) {
if (dagIndex != 1 && interJobSleepTimeout > 0) {
try {
LOG.info("Sleeping between jobs, sleepInterval=" + (interJobSleepTimeout / 1000));
Thread.sleep(interJobSleepTimeout);
} catch (InterruptedException e) {
LOG.info("Main thread interrupted. Breaking out of job loop");
break;
}
}
String inputPath = inputPaths.get(dagIndex - 1);
String outputPath = outputPaths.get(dagIndex - 1);
if (fs.exists(new Path(outputPath))) {
throw new FileAlreadyExistsException("Output directory " + outputPath + " already exists");
}
LOG.info("Running OrderedWordCount DAG" + ", dagIndex=" + dagIndex + ", inputPath=" + inputPath + ", outputPath=" + outputPath);
Map<String, LocalResource> localResources = new TreeMap<String, LocalResource>();
DAG dag = instance.createDAG(fs, tezConf, localResources, stagingDir, dagIndex, inputPath, outputPath, generateSplitsInClient, useMRSettings, intermediateNumReduceTasks, maxDataLengthThroughIPC, exceedDataLimit);
String callerType = "TestOrderedWordCount";
String callerId = tezSession.getAppMasterApplicationId() == null ? ("UnknownApp_" + System.currentTimeMillis() + dagIndex) : (tezSession.getAppMasterApplicationId().toString() + "_" + dagIndex);
dag.setCallerContext(CallerContext.create("Tez", callerId, callerType, "TestOrderedWordCount Job"));
boolean doPreWarm = dagIndex == 1 && useTezSession && conf.getBoolean("PRE_WARM_SESSION", true);
int preWarmNumContainers = 0;
if (doPreWarm) {
preWarmNumContainers = conf.getInt("PRE_WARM_NUM_CONTAINERS", 0);
if (preWarmNumContainers <= 0) {
doPreWarm = false;
}
}
if (doPreWarm) {
LOG.info("Pre-warming Session");
PreWarmVertex preWarmVertex = PreWarmVertex.create("PreWarm", preWarmNumContainers, dag.getVertex("initialmap").getTaskResource());
preWarmVertex.addTaskLocalFiles(dag.getVertex("initialmap").getTaskLocalFiles());
preWarmVertex.setTaskEnvironment(dag.getVertex("initialmap").getTaskEnvironment());
preWarmVertex.setTaskLaunchCmdOpts(dag.getVertex("initialmap").getTaskLaunchCmdOpts());
tezSession.preWarm(preWarmVertex);
}
if (useTezSession) {
LOG.info("Waiting for TezSession to get into ready state");
waitForTezSessionReady(tezSession);
LOG.info("Submitting DAG to Tez Session, dagIndex=" + dagIndex);
dagClient = tezSession.submitDAG(dag);
LOG.info("Submitted DAG to Tez Session, dagIndex=" + dagIndex);
} else {
LOG.info("Submitting DAG as a new Tez Application");
dagClient = tezSession.submitDAG(dag);
}
while (true) {
dagStatus = dagClient.getDAGStatus(statusGetOpts);
if (dagStatus.getState() == DAGStatus.State.RUNNING || dagStatus.getState() == DAGStatus.State.SUCCEEDED || dagStatus.getState() == DAGStatus.State.FAILED || dagStatus.getState() == DAGStatus.State.KILLED || dagStatus.getState() == DAGStatus.State.ERROR) {
break;
}
try {
Thread.sleep(500);
} catch (InterruptedException e) {
// continue;
}
}
while (dagStatus.getState() != DAGStatus.State.SUCCEEDED && dagStatus.getState() != DAGStatus.State.FAILED && dagStatus.getState() != DAGStatus.State.KILLED && dagStatus.getState() != DAGStatus.State.ERROR) {
if (dagStatus.getState() == DAGStatus.State.RUNNING) {
ExampleDriver.printDAGStatus(dagClient, vNames);
}
try {
try {
Thread.sleep(1000);
} catch (InterruptedException e) {
// continue;
}
dagStatus = dagClient.getDAGStatus(statusGetOpts);
} catch (TezException e) {
LOG.error("Failed to get application progress. Exiting");
return -1;
}
}
ExampleDriver.printDAGStatus(dagClient, vNames, true, true);
LOG.info("DAG " + dagIndex + " completed. " + "FinalState=" + dagStatus.getState());
if (dagStatus.getState() != DAGStatus.State.SUCCEEDED) {
LOG.info("DAG " + dagIndex + " diagnostics: " + dagStatus.getDiagnostics());
}
}
} catch (Exception e) {
LOG.error("Error occurred when submitting/running DAGs", e);
throw e;
} finally {
if (!retainStagingDir) {
pathFs.delete(stagingDir, true);
}
LOG.info("Shutting down session");
tezSession.stop();
}
if (!useTezSession) {
ExampleDriver.printDAGStatus(dagClient, vNames);
LOG.info("Application completed. " + "FinalState=" + dagStatus.getState());
}
return dagStatus.getState() == DAGStatus.State.SUCCEEDED ? 0 : 1;
}
use of org.apache.tez.dag.api.client.StatusGetOpts in project tez by apache.
the class ExampleDriver method printDAGStatus.
public static void printDAGStatus(DAGClient dagClient, String[] vertexNames, boolean displayDAGCounters, boolean displayVertexCounters) throws IOException, TezException {
Set<StatusGetOpts> opts = EnumSet.of(StatusGetOpts.GET_COUNTERS);
DAGStatus dagStatus = dagClient.getDAGStatus((displayDAGCounters ? opts : null));
Progress progress = dagStatus.getDAGProgress();
double vProgressFloat = 0.0f;
if (progress != null) {
System.out.println("");
System.out.println("DAG: State: " + dagStatus.getState() + " Progress: " + (progress.getTotalTaskCount() < 0 ? formatter.format(0.0f) : formatter.format((double) (progress.getSucceededTaskCount()) / progress.getTotalTaskCount())));
for (String vertexName : vertexNames) {
VertexStatus vStatus = dagClient.getVertexStatus(vertexName, (displayVertexCounters ? opts : null));
if (vStatus == null) {
System.out.println("Could not retrieve status for vertex: " + vertexName);
continue;
}
Progress vProgress = vStatus.getProgress();
if (vProgress != null) {
vProgressFloat = 0.0f;
if (vProgress.getTotalTaskCount() == 0) {
vProgressFloat = 1.0f;
} else if (vProgress.getTotalTaskCount() > 0) {
vProgressFloat = (double) vProgress.getSucceededTaskCount() / vProgress.getTotalTaskCount();
}
System.out.println("VertexStatus:" + " VertexName: " + (vertexName.equals("ivertex1") ? "intermediate-reducer" : vertexName) + " Progress: " + formatter.format(vProgressFloat));
}
if (displayVertexCounters) {
TezCounters counters = vStatus.getVertexCounters();
if (counters != null) {
System.out.println("Vertex Counters for " + vertexName + ": " + counters);
}
}
}
}
if (displayDAGCounters) {
TezCounters counters = dagStatus.getDAGCounters();
if (counters != null) {
System.out.println("DAG Counters: " + counters);
}
}
}
use of org.apache.tez.dag.api.client.StatusGetOpts in project tez by apache.
the class JoinValidate method runJob.
@Override
protected int runJob(String[] args, TezConfiguration tezConf, TezClient tezClient) throws Exception {
LOG.info("Running JoinValidate");
String lhsDir = args[0];
String rhsDir = args[1];
int numPartitions = 1;
if (args.length == 3) {
numPartitions = Integer.parseInt(args[2]);
}
if (numPartitions <= 0) {
System.err.println("NumPartitions must be > 0");
return 4;
}
Path lhsPath = new Path(lhsDir);
Path rhsPath = new Path(rhsDir);
DAG dag = createDag(tezConf, lhsPath, rhsPath, numPartitions);
tezClient.waitTillReady();
DAGClient dagClient = tezClient.submitDAG(dag);
Set<StatusGetOpts> getOpts = Sets.newHashSet();
if (isCountersLog()) {
getOpts.add(StatusGetOpts.GET_COUNTERS);
}
DAGStatus dagStatus = dagClient.waitForCompletionWithStatusUpdates(getOpts);
if (dagStatus.getState() != DAGStatus.State.SUCCEEDED) {
LOG.info("DAG diagnostics: " + dagStatus.getDiagnostics());
return -1;
} else {
dagStatus = dagClient.getDAGStatus(Sets.newHashSet(StatusGetOpts.GET_COUNTERS));
TezCounter counter = dagStatus.getDAGCounters().findCounter(COUNTER_GROUP_NAME, MISSING_KEY_COUNTER_NAME);
if (counter == null) {
LOG.info("Unable to determing equality");
return -2;
} else {
if (counter.getValue() != 0) {
LOG.info("Validate failed. The two sides are not equivalent");
return -3;
} else {
LOG.info("Validation successful. The two sides are equivalent");
return 0;
}
}
}
}
use of org.apache.tez.dag.api.client.StatusGetOpts in project tez by apache.
the class TezExampleBase method runDag.
/**
* @param dag the dag to execute
* @param printCounters whether to print counters or not
* @param logger the logger to use while printing diagnostics
* @return Zero indicates success, non-zero indicates failure
* @throws TezException
* @throws InterruptedException
* @throws IOException
*/
public int runDag(DAG dag, boolean printCounters, Logger logger) throws TezException, InterruptedException, IOException {
tezClientInternal.waitTillReady();
CallerContext callerContext = CallerContext.create("TezExamples", "Tez Example DAG: " + dag.getName());
ApplicationId appId = tezClientInternal.getAppMasterApplicationId();
if (hadoopShim == null) {
Configuration conf = (getConf() == null ? new Configuration(false) : getConf());
hadoopShim = new HadoopShimsLoader(conf).getHadoopShim();
}
if (appId != null) {
TezUtilsInternal.setHadoopCallerContext(hadoopShim, appId);
callerContext.setCallerIdAndType(appId.toString(), "TezExampleApplication");
}
dag.setCallerContext(callerContext);
DAGClient dagClient = tezClientInternal.submitDAG(dag);
Set<StatusGetOpts> getOpts = Sets.newHashSet();
if (printCounters) {
getOpts.add(StatusGetOpts.GET_COUNTERS);
}
DAGStatus dagStatus;
dagStatus = dagClient.waitForCompletionWithStatusUpdates(getOpts);
if (dagStatus.getState() != DAGStatus.State.SUCCEEDED) {
logger.info("DAG diagnostics: " + dagStatus.getDiagnostics());
return -1;
}
return 0;
}
use of org.apache.tez.dag.api.client.StatusGetOpts in project hive by apache.
the class FSCountersSummary method print.
@Override
public void print(SessionState.LogHelper console) {
console.printInfo("FileSystem Counters Summary");
SortedSet<String> keys = new TreeSet<>(progressMap.keySet());
Set<StatusGetOpts> statusOptions = Collections.singleton(StatusGetOpts.GET_COUNTERS);
// as well. If not, we need a way to get all the schemes that are accessed by the tez task/llap.
for (FileSystem.Statistics statistics : FileSystem.getAllStatistics()) {
final String scheme = statistics.getScheme().toUpperCase();
console.printInfo("");
console.printInfo("Scheme: " + scheme);
console.printInfo(SEPARATOR);
console.printInfo(HEADER);
console.printInfo(SEPARATOR);
for (String vertexName : keys) {
TezCounters vertexCounters = vertexCounters(statusOptions, vertexName);
if (vertexCounters != null) {
console.printInfo(summary(scheme, vertexName, vertexCounters));
}
}
console.printInfo(SEPARATOR);
}
}
Aggregations