List of usage examples for org.apache.hadoop.mapred RunningJob getJobID
@Deprecated
public String getJobID();
From source file:org.apache.pig.backend.hadoop.executionengine.mapreduceExec.MapReduceLauncher.java
License:Apache License
/** * Submit a Pig job to hadoop.//from w ww . j a v a 2s .com * * @param mapFuncs * a list of map functions to apply to the inputs. The cardinality of the list should * be the same as input's cardinality. * @param groupFuncs * a list of grouping functions to apply to the inputs. The cardinality of the list * should be the same as input's cardinality. * @param reduceFunc * the reduce function. * @param mapTasks * the number of map tasks to use. * @param reduceTasks * the number of reduce tasks to use. * @param input * a list of inputs * @param output * the path of the output. * @return an indicator of success or failure. * @throws IOException */ public boolean launchPig(POMapreduce pom) throws IOException { JobConf conf = new JobConf(config); setJobProperties(conf, pom); Properties properties = pom.pigContext.getProperties(); ConfigurationValidator.validatePigProperties(properties); String jobName = properties.getProperty(PigContext.JOB_NAME); conf.setJobName(jobName); boolean success = false; List<String> funcs = new ArrayList<String>(); if (pom.toMap != null) { for (EvalSpec es : pom.toMap) funcs.addAll(es.getFuncs()); } if (pom.groupFuncs != null) { for (EvalSpec es : pom.groupFuncs) funcs.addAll(es.getFuncs()); } if (pom.toReduce != null) { funcs.addAll(pom.toReduce.getFuncs()); } // create jobs.jar locally and pass it to hadoop File submitJarFile = File.createTempFile("Job", ".jar"); try { FileOutputStream fos = new FileOutputStream(submitJarFile); JarManager.createJar(fos, funcs, null, pom.pigContext); log.debug("Job jar size = " + submitJarFile.length()); conf.setJar(submitJarFile.getPath()); String user = System.getProperty("user.name"); conf.setUser(user != null ? user : "Pigster"); conf.set("pig.spill.size.threshold", properties.getProperty("pig.spill.size.threshold")); conf.set("pig.spill.gc.activation.size", properties.getProperty("pig.spill.gc.activation.size")); if (pom.reduceParallelism != -1) { conf.setNumReduceTasks(pom.reduceParallelism); } if (pom.toMap != null) { conf.set("pig.mapFuncs", ObjectSerializer.serialize(pom.toMap)); } if (pom.toCombine != null) { conf.set("pig.combineFunc", ObjectSerializer.serialize(pom.toCombine)); // this is to make sure that combiner is only called once // since we can't handle no combine or multiple combines conf.setCombineOnceOnly(true); } if (pom.groupFuncs != null) { conf.set("pig.groupFuncs", ObjectSerializer.serialize(pom.groupFuncs)); } if (pom.toReduce != null) { conf.set("pig.reduceFunc", ObjectSerializer.serialize(pom.toReduce)); } if (pom.toSplit != null) { conf.set("pig.splitSpec", ObjectSerializer.serialize(pom.toSplit)); } if (pom.pigContext != null) { conf.set("pig.pigContext", ObjectSerializer.serialize(pom.pigContext)); } conf.setMapRunnerClass(PigMapReduce.class); if (pom.toCombine != null) { conf.setCombinerClass(PigCombine.class); //conf.setCombinerClass(PigMapReduce.class); } if (pom.quantilesFile != null) { conf.set("pig.quantilesFile", pom.quantilesFile); } else { // this is not a sort job - can use byte comparison to speed up processing conf.setOutputKeyComparatorClass(PigWritableComparator.class); } if (pom.partitionFunction != null) { conf.setPartitionerClass(SortPartitioner.class); } conf.setReducerClass(PigMapReduce.class); conf.setInputFormat(PigInputFormat.class); conf.setOutputFormat(PigOutputFormat.class); // not used starting with 0.15 conf.setInputKeyClass(Text.class); // not used starting with 0.15 conf.setInputValueClass(Tuple.class); conf.setOutputKeyClass(Tuple.class); if (pom.userComparator != null) { conf.setOutputKeyComparatorClass(pom.userComparator); } conf.setOutputValueClass(IndexedTuple.class); conf.set("pig.inputs", ObjectSerializer.serialize(pom.inputFileSpecs)); conf.setOutputPath(new Path(pom.outputFileSpec.getFileName())); conf.set("pig.storeFunc", ObjectSerializer.serialize(pom.outputFileSpec.getFuncSpec())); // Setup the DistributedCache for this job setupDistributedCache(pom.pigContext, conf, pom.properties, "pig.streaming.ship.files", true); setupDistributedCache(pom.pigContext, conf, pom.properties, "pig.streaming.cache.files", false); // Setup the logs directory for this job String jobOutputFileName = pom.pigContext.getJobOutputFile(); if (jobOutputFileName != null && jobOutputFileName.length() > 0) { Path jobOutputFile = new Path(pom.pigContext.getJobOutputFile()); conf.set("pig.output.dir", jobOutputFile.getParent().toString()); conf.set("pig.streaming.log.dir", new Path(jobOutputFile, LOG_DIR).toString()); } // // Now, actually submit the job (using the submit name) // JobClient jobClient = execEngine.getJobClient(); RunningJob status = jobClient.submitJob(conf); log.debug("submitted job: " + status.getJobID()); long sleepTime = 1000; double lastQueryProgress = -1.0; int lastJobsQueued = -1; double lastMapProgress = -1.0; double lastReduceProgress = -1.0; while (true) { try { Thread.sleep(sleepTime); } catch (Exception e) { } if (status.isComplete()) { success = status.isSuccessful(); if (log.isDebugEnabled()) { StringBuilder sb = new StringBuilder(); sb.append("Job finished "); sb.append((success ? "" : "un")); sb.append("successfully"); log.debug(sb.toString()); } if (success) { mrJobNumber++; } double queryProgress = ((double) mrJobNumber) / ((double) numMRJobs); if (queryProgress > lastQueryProgress) { if (log.isInfoEnabled()) { StringBuilder sbProgress = new StringBuilder(); sbProgress.append("Pig progress = "); sbProgress.append(((int) (queryProgress * 100))); sbProgress.append("%"); log.info(sbProgress.toString()); } lastQueryProgress = queryProgress; } break; } else // still running { double mapProgress = status.mapProgress(); double reduceProgress = status.reduceProgress(); if (lastMapProgress != mapProgress || lastReduceProgress != reduceProgress) { if (log.isDebugEnabled()) { StringBuilder sbProgress = new StringBuilder(); sbProgress.append("Hadoop job progress: Map="); sbProgress.append((int) (mapProgress * 100)); sbProgress.append("% Reduce="); sbProgress.append((int) (reduceProgress * 100)); sbProgress.append("%"); log.debug(sbProgress.toString()); } lastMapProgress = mapProgress; lastReduceProgress = reduceProgress; } double numJobsCompleted = mrJobNumber; double thisJobProgress = (mapProgress + reduceProgress) / 2.0; double queryProgress = (numJobsCompleted + thisJobProgress) / ((double) numMRJobs); if (queryProgress > lastQueryProgress) { if (log.isInfoEnabled()) { StringBuilder sbProgress = new StringBuilder(); sbProgress.append("Pig progress = "); sbProgress.append(((int) (queryProgress * 100))); sbProgress.append("%"); log.info(sbProgress.toString()); } lastQueryProgress = queryProgress; } } } // bug 1030028: if the input file is empty; hadoop doesn't create the output file! Path outputFile = conf.getOutputPath(); String outputName = outputFile.getName(); int colon = outputName.indexOf(':'); if (colon != -1) { outputFile = new Path(outputFile.getParent(), outputName.substring(0, colon)); } try { ElementDescriptor descriptor = ((HDataStorage) (pom.pigContext.getDfs())) .asElement(outputFile.toString()); if (success && !descriptor.exists()) { // create an empty output file PigFile f = new PigFile(outputFile.toString(), false); f.store(BagFactory.getInstance().newDefaultBag(), new PigStorage(), pom.pigContext); } } catch (DataStorageException e) { throw WrappedIOException.wrap("Failed to obtain descriptor for " + outputFile.toString(), e); } if (!success) { // go find the error messages getErrorMessages(jobClient.getMapTaskReports(status.getJobID()), "map"); getErrorMessages(jobClient.getReduceTaskReports(status.getJobID()), "reduce"); } else { long timeSpent = 0; // NOTE: this call is crashing due to a bug in Hadoop; the bug is known and the patch has not been applied yet. TaskReport[] mapReports = jobClient.getMapTaskReports(status.getJobID()); TaskReport[] reduceReports = jobClient.getReduceTaskReports(status.getJobID()); for (TaskReport r : mapReports) { timeSpent += (r.getFinishTime() - r.getStartTime()); } for (TaskReport r : reduceReports) { timeSpent += (r.getFinishTime() - r.getStartTime()); } totalHadoopTimeSpent += timeSpent; } } catch (Exception e) { // Do we need different handling for different exceptions e.printStackTrace(); throw WrappedIOException.wrap(e); } finally { submitJarFile.delete(); } return success; }