List of usage examples for org.apache.hadoop.security Credentials addAll
public void addAll(Credentials other)
From source file:com.cloudera.hue.CredentialsMerger.java
License:Apache License
/** * Merge several credentials files into one. Give the desired output file * first, followed by all of the input files. * * <p>File formats are tried in this order: TokenStorageFile, urlEncodedString. * </p>/*from ww w . j a v a 2 s .c o m*/ * * @param args <out> <in1> ... * @throws IOException in the event of an error reading or writing files. */ public static void main(String[] args) throws IOException { if (args.length < 2) { printUsage(); System.exit(1); } Path outputFile = new Path("file://" + new File(args[0]).getAbsolutePath()); Configuration conf = new Configuration(); Credentials credentials = new Credentials(); for (int i = 1; i < args.length; i++) { try { Credentials singleFileCredentials = Credentials .readTokenStorageFile(new Path("file://" + new File(args[i]).getAbsolutePath()), conf); credentials.addAll(singleFileCredentials); } catch (IOException e) { BufferedReader reader = new BufferedReader(new FileReader(args[i])); try { // Retry to read the token with an encodedUrl format Token<?> token = new Token(); String encodedtoken = reader.readLine(); token.decodeFromUrlString(encodedtoken); credentials.addToken(new Text(args[i]), token); } finally { reader.close(); } } } credentials.writeTokenStorageFile(outputFile, conf); }
From source file:com.cloudera.recordservice.pig.HCatRSLoader.java
License:Apache License
@Override public void setLocation(String location, Job job) throws IOException { HCatContext.INSTANCE.setConf(job.getConfiguration()).getConf().get() .setBoolean(HCatConstants.HCAT_DATA_TINY_SMALL_INT_PROMOTION, true); UDFContext udfContext = UDFContext.getUDFContext(); Properties udfProps = udfContext.getUDFProperties(this.getClass(), new String[] { signature }); job.getConfiguration().set(INNER_SIGNATURE, INNER_SIGNATURE_PREFIX + "_" + signature); RequiredFieldList requiredFieldsInfo = (RequiredFieldList) udfProps.get(PRUNE_PROJECTION_INFO); // get partitionFilterString stored in the UDFContext - it would have // been stored there by an earlier call to setPartitionFilter // call setInput on HCatInputFormat only in the frontend because internally // it makes calls to the hcat server - we don't want these to happen in // the backend // in the hadoop front end mapred.task.id property will not be set in // the Configuration if (udfProps.containsKey(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET)) { for (Enumeration<Object> emr = udfProps.keys(); emr.hasMoreElements();) { PigHCatUtil.getConfigFromUDFProperties(udfProps, job.getConfiguration(), emr.nextElement().toString()); }// w w w . j a va 2 s.c om if (!HCatUtil.checkJobContextIfRunningFromBackend(job)) { //Combine credentials and credentials from job takes precedence for freshness Credentials crd = jobCredentials.get(INNER_SIGNATURE_PREFIX + "_" + signature); job.getCredentials().addAll(crd); } } else { Job clone = new Job(job.getConfiguration()); HCatRSInputFormat.setInput(job, location, getPartitionFilterString()); InputJobInfo inputJobInfo = (InputJobInfo) HCatRSUtil .deserialize(job.getConfiguration().get(HCatConstants.HCAT_KEY_JOB_INFO)); // TODO: Add back special cases call when I find out where the code has moved. addSpecialCasesParametersForHCatLoader(job.getConfiguration(), inputJobInfo.getTableInfo()); // We will store all the new /changed properties in the job in the // udf context, so the the HCatInputFormat.setInput method need not //be called many times. for (Entry<String, String> keyValue : job.getConfiguration()) { String oldValue = clone.getConfiguration().getRaw(keyValue.getKey()); if ((oldValue == null) || (keyValue.getValue().equals(oldValue) == false)) { udfProps.put(keyValue.getKey(), keyValue.getValue()); } } udfProps.put(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET, true); //Store credentials in a private hash map and not the udf context to // make sure they are not public. Credentials crd = new Credentials(); crd.addAll(job.getCredentials()); jobCredentials.put(INNER_SIGNATURE_PREFIX + "_" + signature, crd); clone.setInputFormatClass(HCatRSInputFormat.class); } // Need to also push projections by calling setOutputSchema on // HCatInputFormat - we have to get the RequiredFields information // from the UdfContext, translate it to an Schema and then pass it // The reason we do this here is because setLocation() is called by // Pig runtime at InputFormat.getSplits() and // InputFormat.createRecordReader() time - we are not sure when // HCatInputFormat needs to know about pruned projections - so doing it // here will ensure we communicate to HCatInputFormat about pruned // projections at getSplits() and createRecordReader() time if (requiredFieldsInfo != null) { // convert to hcatschema and pass to HCatInputFormat try { outputSchema = phutil.getHCatSchema(requiredFieldsInfo.getFields(), signature, this.getClass()); HCatRSInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } else { // else - this means pig's optimizer never invoked the pushProjection // method - so we need all fields and hence we should not call the // setOutputSchema on HCatInputFormat if (HCatUtil.checkJobContextIfRunningFromBackend(job)) { try { HCatSchema hcatTableSchema = (HCatSchema) udfProps.get(HCatConstants.HCAT_TABLE_SCHEMA); outputSchema = hcatTableSchema; HCatRSInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } } if (LOG.isDebugEnabled()) { LOG.debug("outputSchema=" + outputSchema); } job.setInputFormatClass(HCatRSInputFormat.class); }
From source file:com.datatorrent.stram.security.StramUserLogin.java
License:Apache License
public static long refreshTokens(long tokenLifeTime, String destinationDir, String destinationFile, final Configuration conf, String hdfsKeyTabFile, final Credentials credentials, final InetSocketAddress rmAddress, final boolean renewRMToken) throws IOException { long expiryTime = System.currentTimeMillis() + tokenLifeTime; //renew tokens final String tokenRenewer = conf.get(YarnConfiguration.RM_PRINCIPAL); if (tokenRenewer == null || tokenRenewer.length() == 0) { throw new IOException("Can't get Master Kerberos principal for the RM to use as renewer"); }//from w w w. j ava 2s .co m FileSystem fs = FileSystem.newInstance(conf); File keyTabFile; try { keyTabFile = FSUtil.copyToLocalFileSystem(fs, destinationDir, destinationFile, hdfsKeyTabFile, conf); } finally { fs.close(); } UserGroupInformation ugi = UserGroupInformation.loginUserFromKeytabAndReturnUGI( UserGroupInformation.getCurrentUser().getUserName(), keyTabFile.getAbsolutePath()); try { ugi.doAs(new PrivilegedExceptionAction<Object>() { @Override public Object run() throws Exception { FileSystem fs1 = FileSystem.newInstance(conf); YarnClient yarnClient = null; if (renewRMToken) { yarnClient = YarnClient.createYarnClient(); yarnClient.init(conf); yarnClient.start(); } Credentials creds = new Credentials(); try { fs1.addDelegationTokens(tokenRenewer, creds); if (renewRMToken) { org.apache.hadoop.yarn.api.records.Token rmDelToken = yarnClient .getRMDelegationToken(new Text(tokenRenewer)); Token<RMDelegationTokenIdentifier> rmToken = ConverterUtils.convertFromYarn(rmDelToken, rmAddress); creds.addToken(rmToken.getService(), rmToken); } } finally { fs1.close(); if (renewRMToken) { yarnClient.stop(); } } credentials.addAll(creds); return null; } }); UserGroupInformation.getCurrentUser().addCredentials(credentials); } catch (InterruptedException e) { LOG.error("Error while renewing tokens ", e); expiryTime = System.currentTimeMillis(); } catch (IOException e) { LOG.error("Error while renewing tokens ", e); expiryTime = System.currentTimeMillis(); } LOG.debug("number of tokens: {}", credentials.getAllTokens().size()); Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.debug("updated token: {}", token); } keyTabFile.delete(); return expiryTime; }
From source file:org.apache.gobblin.yarn.GobblinYarnAppLauncher.java
License:Apache License
private void setupSecurityTokens(ContainerLaunchContext containerLaunchContext) throws IOException { Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); // Pass on the credentials from the hadoop token file if present. // The value in the token file takes precedence. if (System.getenv(HADOOP_TOKEN_FILE_LOCATION) != null) { Credentials tokenFileCredentials = Credentials .readTokenStorageFile(new File(System.getenv(HADOOP_TOKEN_FILE_LOCATION)), new Configuration()); credentials.addAll(tokenFileCredentials); }//w w w.jav a 2s. c o m String tokenRenewer = this.yarnConfiguration.get(YarnConfiguration.RM_PRINCIPAL); if (tokenRenewer == null || tokenRenewer.length() == 0) { throw new IOException("Failed to get master Kerberos principal for the RM to use as renewer"); } // For now, only getting tokens for the default file-system. Token<?> tokens[] = this.fs.addDelegationTokens(tokenRenewer, credentials); if (tokens != null) { for (Token<?> token : tokens) { LOGGER.info("Got delegation token for " + this.fs.getUri() + "; " + token); } } Closer closer = Closer.create(); try { DataOutputBuffer dataOutputBuffer = closer.register(new DataOutputBuffer()); credentials.writeTokenStorageToStream(dataOutputBuffer); ByteBuffer fsTokens = ByteBuffer.wrap(dataOutputBuffer.getData(), 0, dataOutputBuffer.getLength()); containerLaunchContext.setTokens(fsTokens); } catch (Throwable t) { throw closer.rethrow(t); } finally { closer.close(); } }
From source file:org.apache.hcatalog.pig.HCatLoader.java
License:Apache License
@Override public void setLocation(String location, Job job) throws IOException { HCatContext.INSTANCE.setConf(job.getConfiguration()).getConf().get() .setBoolean(HCatConstants.HCAT_DATA_TINY_SMALL_INT_PROMOTION, true); UDFContext udfContext = UDFContext.getUDFContext(); Properties udfProps = udfContext.getUDFProperties(this.getClass(), new String[] { signature }); job.getConfiguration().set(INNER_SIGNATURE, INNER_SIGNATURE_PREFIX + "_" + signature); Pair<String, String> dbTablePair = PigHCatUtil.getDBTableNames(location); dbName = dbTablePair.first;//from w ww. j a v a 2 s . c o m tableName = dbTablePair.second; RequiredFieldList requiredFieldsInfo = (RequiredFieldList) udfProps.get(PRUNE_PROJECTION_INFO); // get partitionFilterString stored in the UDFContext - it would have // been stored there by an earlier call to setPartitionFilter // call setInput on HCatInputFormat only in the frontend because internally // it makes calls to the hcat server - we don't want these to happen in // the backend // in the hadoop front end mapred.task.id property will not be set in // the Configuration if (udfProps.containsKey(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET)) { for (Enumeration<Object> emr = udfProps.keys(); emr.hasMoreElements();) { PigHCatUtil.getConfigFromUDFProperties(udfProps, job.getConfiguration(), emr.nextElement().toString()); } if (!HCatUtil.checkJobContextIfRunningFromBackend(job)) { //Combine credentials and credentials from job takes precedence for freshness Credentials crd = jobCredentials.get(INNER_SIGNATURE_PREFIX + "_" + signature); crd.addAll(job.getCredentials()); job.getCredentials().addAll(crd); } } else { Job clone = new Job(job.getConfiguration()); HCatInputFormat.setInput(job, dbName, tableName).setFilter(getPartitionFilterString()); // We will store all the new /changed properties in the job in the // udf context, so the the HCatInputFormat.setInput method need not //be called many times. for (Entry<String, String> keyValue : job.getConfiguration()) { String oldValue = clone.getConfiguration().getRaw(keyValue.getKey()); if ((oldValue == null) || (keyValue.getValue().equals(oldValue) == false)) { udfProps.put(keyValue.getKey(), keyValue.getValue()); } } udfProps.put(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET, true); //Store credentials in a private hash map and not the udf context to // make sure they are not public. Credentials crd = new Credentials(); crd.addAll(job.getCredentials()); jobCredentials.put(INNER_SIGNATURE_PREFIX + "_" + signature, crd); } // Need to also push projections by calling setOutputSchema on // HCatInputFormat - we have to get the RequiredFields information // from the UdfContext, translate it to an Schema and then pass it // The reason we do this here is because setLocation() is called by // Pig runtime at InputFormat.getSplits() and // InputFormat.createRecordReader() time - we are not sure when // HCatInputFormat needs to know about pruned projections - so doing it // here will ensure we communicate to HCatInputFormat about pruned // projections at getSplits() and createRecordReader() time if (requiredFieldsInfo != null) { // convert to hcatschema and pass to HCatInputFormat try { outputSchema = phutil.getHCatSchema(requiredFieldsInfo.getFields(), signature, this.getClass()); HCatInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } else { // else - this means pig's optimizer never invoked the pushProjection // method - so we need all fields and hence we should not call the // setOutputSchema on HCatInputFormat if (HCatUtil.checkJobContextIfRunningFromBackend(job)) { try { HCatSchema hcatTableSchema = (HCatSchema) udfProps.get(HCatConstants.HCAT_TABLE_SCHEMA); outputSchema = hcatTableSchema; HCatInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } } }
From source file:org.apache.hive.hcatalog.pig.HCatLoader.java
License:Apache License
@Override public void setLocation(String location, Job job) throws IOException { HCatContext.INSTANCE.setConf(job.getConfiguration()).getConf().get() .setBoolean(HCatConstants.HCAT_DATA_TINY_SMALL_INT_PROMOTION, true); UDFContext udfContext = UDFContext.getUDFContext(); Properties udfProps = udfContext.getUDFProperties(this.getClass(), new String[] { signature }); job.getConfiguration().set(INNER_SIGNATURE, INNER_SIGNATURE_PREFIX + "_" + signature); Pair<String, String> dbTablePair = PigHCatUtil.getDBTableNames(location); dbName = dbTablePair.first;// w w w.j av a 2s. co m tableName = dbTablePair.second; RequiredFieldList requiredFieldsInfo = (RequiredFieldList) udfProps.get(PRUNE_PROJECTION_INFO); // get partitionFilterString stored in the UDFContext - it would have // been stored there by an earlier call to setPartitionFilter // call setInput on HCatInputFormat only in the frontend because internally // it makes calls to the hcat server - we don't want these to happen in // the backend // in the hadoop front end mapred.task.id property will not be set in // the Configuration if (udfProps.containsKey(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET)) { for (Enumeration<Object> emr = udfProps.keys(); emr.hasMoreElements();) { PigHCatUtil.getConfigFromUDFProperties(udfProps, job.getConfiguration(), emr.nextElement().toString()); } if (!HCatUtil.checkJobContextIfRunningFromBackend(job)) { //Combine credentials and credentials from job takes precedence for freshness Credentials crd = jobCredentials.get(INNER_SIGNATURE_PREFIX + "_" + signature); job.getCredentials().addAll(crd); } } else { Job clone = new Job(job.getConfiguration()); HCatInputFormat.setInput(job, dbName, tableName, getPartitionFilterString()); InputJobInfo inputJobInfo = (InputJobInfo) HCatUtil .deserialize(job.getConfiguration().get(HCatConstants.HCAT_KEY_JOB_INFO)); SpecialCases.addSpecialCasesParametersForHCatLoader(job.getConfiguration(), inputJobInfo.getTableInfo()); // We will store all the new /changed properties in the job in the // udf context, so the the HCatInputFormat.setInput method need not //be called many times. for (Entry<String, String> keyValue : job.getConfiguration()) { String oldValue = clone.getConfiguration().getRaw(keyValue.getKey()); if ((oldValue == null) || (keyValue.getValue().equals(oldValue) == false)) { udfProps.put(keyValue.getKey(), keyValue.getValue()); } } udfProps.put(HCatConstants.HCAT_PIG_LOADER_LOCATION_SET, true); //Store credentials in a private hash map and not the udf context to // make sure they are not public. Credentials crd = new Credentials(); crd.addAll(job.getCredentials()); jobCredentials.put(INNER_SIGNATURE_PREFIX + "_" + signature, crd); } // Need to also push projections by calling setOutputSchema on // HCatInputFormat - we have to get the RequiredFields information // from the UdfContext, translate it to an Schema and then pass it // The reason we do this here is because setLocation() is called by // Pig runtime at InputFormat.getSplits() and // InputFormat.createRecordReader() time - we are not sure when // HCatInputFormat needs to know about pruned projections - so doing it // here will ensure we communicate to HCatInputFormat about pruned // projections at getSplits() and createRecordReader() time if (requiredFieldsInfo != null) { // convert to hcatschema and pass to HCatInputFormat try { //push down projections to columnar store works for RCFile and ORCFile ArrayList<Integer> list = new ArrayList<Integer>(requiredFieldsInfo.getFields().size()); for (RequiredField rf : requiredFieldsInfo.getFields()) { list.add(rf.getIndex()); } ColumnProjectionUtils.setReadColumns(job.getConfiguration(), list); outputSchema = phutil.getHCatSchema(requiredFieldsInfo.getFields(), signature, this.getClass()); HCatInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } else { // else - this means pig's optimizer never invoked the pushProjection // method - so we need all fields and hence we should not call the // setOutputSchema on HCatInputFormat ColumnProjectionUtils.setReadAllColumns(job.getConfiguration()); if (HCatUtil.checkJobContextIfRunningFromBackend(job)) { try { HCatSchema hcatTableSchema = (HCatSchema) udfProps.get(HCatConstants.HCAT_TABLE_SCHEMA); outputSchema = hcatTableSchema; HCatInputFormat.setOutputSchema(job, outputSchema); } catch (Exception e) { throw new IOException(e); } } } if (LOG.isDebugEnabled()) { LOG.debug("outputSchema=" + outputSchema); } }
From source file:org.apache.pig.backend.hadoop.executionengine.tez.util.SecurityHelper.java
License:Apache License
@SuppressWarnings("unchecked") private static void readTokensFromFiles(Configuration conf, Credentials credentials) throws IOException { // add tokens and secrets coming from a token storage file String binaryTokenFilename = conf.get(MRConfiguration.JOB_CREDENTIALS_BINARY); if (binaryTokenFilename != null) { Credentials binary = Credentials.readTokenStorageFile(new Path("file:///" + binaryTokenFilename), conf); credentials.addAll(binary); }// w ww . j a va2s. co m // add secret keys coming from a json file String tokensFileName = conf.get(MRConfiguration.JOB_CREDENTIALS_JSON); if (tokensFileName != null) { LOG.info("loading user's secret keys from " + tokensFileName); String localFileName = new Path(tokensFileName).toUri().getPath(); boolean json_error = false; try { // read JSON ObjectMapper mapper = new ObjectMapper(); Map<String, String> nm = mapper.readValue(new File(localFileName), Map.class); for (Map.Entry<String, String> ent : nm.entrySet()) { credentials.addSecretKey(new Text(ent.getKey()), ent.getValue().getBytes(Charsets.UTF_8)); } } catch (JsonMappingException e) { json_error = true; } catch (JsonParseException e) { json_error = true; } if (json_error) LOG.warn("couldn't parse Token Cache JSON file with user secret keys"); } }
From source file:org.apache.tez.client.TezClientUtils.java
License:Apache License
/** * Create an ApplicationSubmissionContext to launch a Tez AM * @param appId Application Id/* w w w .j av a 2 s . c o m*/ * @param dag DAG to be submitted * @param amName Name for the application * @param amConfig AM Configuration * @param tezJarResources Resources to be used by the AM * @param sessionCreds the credential object which will be populated with session specific * @param historyACLPolicyManager * @return an ApplicationSubmissionContext to launch a Tez AM * @throws IOException * @throws YarnException */ @Private @VisibleForTesting public static ApplicationSubmissionContext createApplicationSubmissionContext(ApplicationId appId, DAG dag, String amName, AMConfiguration amConfig, Map<String, LocalResource> tezJarResources, Credentials sessionCreds, boolean tezLrsAsArchive, TezApiVersionInfo apiVersionInfo, HistoryACLPolicyManager historyACLPolicyManager) throws IOException, YarnException { Preconditions.checkNotNull(sessionCreds); TezConfiguration conf = amConfig.getTezConfiguration(); FileSystem fs = TezClientUtils.ensureStagingDirExists(conf, TezCommonUtils.getTezBaseStagingPath(conf)); String strAppId = appId.toString(); Path tezSysStagingPath = TezCommonUtils.createTezSystemStagingPath(conf, strAppId); Path binaryConfPath = TezCommonUtils.getTezConfStagingPath(tezSysStagingPath); binaryConfPath = fs.makeQualified(binaryConfPath); // Setup resource requirements Resource capability = Records.newRecord(Resource.class); capability.setMemory(amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_AM_RESOURCE_MEMORY_MB, TezConfiguration.TEZ_AM_RESOURCE_MEMORY_MB_DEFAULT)); capability.setVirtualCores(amConfig.getTezConfiguration().getInt( TezConfiguration.TEZ_AM_RESOURCE_CPU_VCORES, TezConfiguration.TEZ_AM_RESOURCE_CPU_VCORES_DEFAULT)); if (LOG.isDebugEnabled()) { LOG.debug("AppMaster capability = " + capability); } // Setup required Credentials for the AM launch. DAG specific credentials // are handled separately. ByteBuffer securityTokens = null; // Setup security tokens Credentials amLaunchCredentials = new Credentials(); if (amConfig.getCredentials() != null) { amLaunchCredentials.addAll(amConfig.getCredentials()); } // Add Staging dir creds to the list of session credentials. TokenCache.obtainTokensForFileSystems(sessionCreds, new Path[] { binaryConfPath }, conf); // Add session specific credentials to the AM credentials. amLaunchCredentials.mergeAll(sessionCreds); DataOutputBuffer dob = new DataOutputBuffer(); amLaunchCredentials.writeTokenStorageToStream(dob); securityTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Setup the command to run the AM List<String> vargs = new ArrayList<String>(8); vargs.add(Environment.JAVA_HOME.$() + "/bin/java"); String amOpts = constructAMLaunchOpts(amConfig.getTezConfiguration(), capability); vargs.add(amOpts); String amLogLevelString = amConfig.getTezConfiguration().get(TezConfiguration.TEZ_AM_LOG_LEVEL, TezConfiguration.TEZ_AM_LOG_LEVEL_DEFAULT); String[] amLogParams = parseLogParams(amLogLevelString); String amLogLevel = amLogParams[0]; maybeAddDefaultLoggingJavaOpts(amLogLevel, vargs); // FIX sun bug mentioned in TEZ-327 vargs.add("-Dsun.nio.ch.bugLevel=''"); vargs.add(TezConstants.TEZ_APPLICATION_MASTER_CLASS); if (dag == null) { vargs.add("--" + TezConstants.TEZ_SESSION_MODE_CLI_OPTION); } vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + ApplicationConstants.STDOUT); vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + ApplicationConstants.STDERR); Vector<String> vargsFinal = new Vector<String>(8); // Final command StringBuilder mergedCommand = new StringBuilder(); for (CharSequence str : vargs) { mergedCommand.append(str).append(" "); } vargsFinal.add(mergedCommand.toString()); if (LOG.isDebugEnabled()) { LOG.debug("Command to launch container for ApplicationMaster is : " + mergedCommand); } Map<String, String> environment = new TreeMap<String, String>(); TezYARNUtils.setupDefaultEnv(environment, conf, TezConfiguration.TEZ_AM_LAUNCH_ENV, TezConfiguration.TEZ_AM_LAUNCH_ENV_DEFAULT, tezLrsAsArchive); addVersionInfoToEnv(environment, apiVersionInfo); addLogParamsToEnv(environment, amLogParams); Map<String, LocalResource> amLocalResources = new TreeMap<String, LocalResource>(); // Not fetching credentials for AMLocalResources. Expect this to be provided via AMCredentials. if (amConfig.getAMLocalResources() != null) { amLocalResources.putAll(amConfig.getAMLocalResources()); } amLocalResources.putAll(tezJarResources); // Setup Session ACLs and update conf as needed Map<String, String> aclConfigs = null; if (historyACLPolicyManager != null) { if (dag == null) { aclConfigs = historyACLPolicyManager.setupSessionACLs(amConfig.getTezConfiguration(), appId); } else { // Non-session mode // As only a single DAG is support, we should combine AM and DAG ACLs under the same // acl management layer aclConfigs = historyACLPolicyManager.setupNonSessionACLs(amConfig.getTezConfiguration(), appId, dag.getDagAccessControls()); } } // emit conf as PB file ConfigurationProto finalConfProto = createFinalConfProtoForApp(amConfig.getTezConfiguration(), aclConfigs); FSDataOutputStream amConfPBOutBinaryStream = null; try { amConfPBOutBinaryStream = TezCommonUtils.createFileForAM(fs, binaryConfPath); finalConfProto.writeTo(amConfPBOutBinaryStream); } finally { if (amConfPBOutBinaryStream != null) { amConfPBOutBinaryStream.close(); } } LocalResource binaryConfLRsrc = TezClientUtils.createLocalResource(fs, binaryConfPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION); amConfig.setBinaryConfLR(binaryConfLRsrc); amLocalResources.put(TezConstants.TEZ_PB_BINARY_CONF_NAME, binaryConfLRsrc); // Create Session Jars definition to be sent to AM as a local resource Path sessionJarsPath = TezCommonUtils.getTezAMJarStagingPath(tezSysStagingPath); FSDataOutputStream sessionJarsPBOutStream = null; try { sessionJarsPBOutStream = TezCommonUtils.createFileForAM(fs, sessionJarsPath); // Write out the initial list of resources which will be available in the AM DAGProtos.PlanLocalResourcesProto amResourceProto; if (amLocalResources != null && !amLocalResources.isEmpty()) { amResourceProto = DagTypeConverters.convertFromLocalResources(amLocalResources); } else { amResourceProto = DAGProtos.PlanLocalResourcesProto.getDefaultInstance(); } amResourceProto.writeDelimitedTo(sessionJarsPBOutStream); } finally { if (sessionJarsPBOutStream != null) { sessionJarsPBOutStream.close(); } } LocalResource sessionJarsPBLRsrc = TezClientUtils.createLocalResource(fs, sessionJarsPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION); amLocalResources.put(TezConstants.TEZ_AM_LOCAL_RESOURCES_PB_FILE_NAME, sessionJarsPBLRsrc); String user = UserGroupInformation.getCurrentUser().getShortUserName(); ACLManager aclManager = new ACLManager(user, amConfig.getTezConfiguration()); Map<ApplicationAccessType, String> acls = aclManager.toYARNACls(); if (dag != null) { DAGPlan dagPB = prepareAndCreateDAGPlan(dag, amConfig, tezJarResources, tezLrsAsArchive, sessionCreds); // emit protobuf DAG file style Path binaryPath = TezCommonUtils.getTezBinPlanStagingPath(tezSysStagingPath); if (LOG.isDebugEnabled()) { LOG.debug("Stage directory information for AppId :" + appId + " tezSysStagingPath :" + tezSysStagingPath + " binaryConfPath :" + binaryConfPath + " sessionJarsPath :" + sessionJarsPath + " binaryPlanPath :" + binaryPath); } FSDataOutputStream dagPBOutBinaryStream = null; try { //binary output dagPBOutBinaryStream = TezCommonUtils.createFileForAM(fs, binaryPath); dagPB.writeTo(dagPBOutBinaryStream); } finally { if (dagPBOutBinaryStream != null) { dagPBOutBinaryStream.close(); } } amLocalResources.put(TezConstants.TEZ_PB_PLAN_BINARY_NAME, TezClientUtils.createLocalResource(fs, binaryPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION)); if (Level.DEBUG.isGreaterOrEqual(Level.toLevel(amLogLevel))) { Path textPath = localizeDagPlanAsText(dagPB, fs, amConfig, strAppId, tezSysStagingPath); amLocalResources.put(TezConstants.TEZ_PB_PLAN_TEXT_NAME, TezClientUtils.createLocalResource(fs, textPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION)); } } // Setup ContainerLaunchContext for AM container ContainerLaunchContext amContainer = ContainerLaunchContext.newInstance(amLocalResources, environment, vargsFinal, null, securityTokens, acls); // Set up the ApplicationSubmissionContext ApplicationSubmissionContext appContext = Records.newRecord(ApplicationSubmissionContext.class); appContext.setApplicationType(TezConstants.TEZ_APPLICATION_TYPE); appContext.setApplicationId(appId); appContext.setResource(capability); if (amConfig.getQueueName() != null) { appContext.setQueue(amConfig.getQueueName()); } appContext.setApplicationName(amName); appContext.setCancelTokensWhenComplete(amConfig.getTezConfiguration().getBoolean( TezConfiguration.TEZ_CANCEL_DELEGATION_TOKENS_ON_COMPLETION, TezConfiguration.TEZ_CANCEL_DELEGATION_TOKENS_ON_COMPLETION_DEFAULT)); appContext.setAMContainerSpec(amContainer); appContext.setMaxAppAttempts(amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_AM_MAX_APP_ATTEMPTS, TezConfiguration.TEZ_AM_MAX_APP_ATTEMPTS_DEFAULT)); return appContext; }
From source file:org.apache.tez.dag.api.DAG.java
License:Apache License
@Private public synchronized DAGPlan createDag(Configuration tezConf, Credentials extraCredentials, Map<String, LocalResource> tezJarResources, LocalResource binaryConfig, boolean tezLrsAsArchive, Map<String, String> additionalConfigs) { verify(true);// ww w. ja v a 2 s. c o m DAGPlan.Builder dagBuilder = DAGPlan.newBuilder(); dagBuilder.setName(this.name); if (this.dagInfo != null && !this.dagInfo.isEmpty()) { dagBuilder.setDagInfo(this.dagInfo); } if (!vertexGroups.isEmpty()) { for (VertexGroup av : vertexGroups) { GroupInfo groupInfo = av.getGroupInfo(); PlanVertexGroupInfo.Builder groupBuilder = PlanVertexGroupInfo.newBuilder(); groupBuilder.setGroupName(groupInfo.getGroupName()); for (Vertex v : groupInfo.getMembers()) { groupBuilder.addGroupMembers(v.getName()); } groupBuilder.addAllOutputs(groupInfo.outputs); for (Map.Entry<String, InputDescriptor> entry : groupInfo.edgeMergedInputs.entrySet()) { groupBuilder.addEdgeMergedInputs( PlanGroupInputEdgeInfo.newBuilder().setDestVertexName(entry.getKey()) .setMergedInput(DagTypeConverters.convertToDAGPlan(entry.getValue()))); } dagBuilder.addVertexGroups(groupBuilder); } } Credentials dagCredentials = new Credentials(); if (extraCredentials != null) { dagCredentials.mergeAll(extraCredentials); } dagCredentials.mergeAll(credentials); if (!commonTaskLocalFiles.isEmpty()) { dagBuilder.addAllLocalResource(DagTypeConverters.convertToDAGPlan(commonTaskLocalFiles)); } Preconditions.checkArgument(topologicalVertexStack.size() == vertices.size(), "size of topologicalVertexStack is:" + topologicalVertexStack.size() + " while size of vertices is:" + vertices.size() + ", make sure they are the same in order to sort the vertices"); while (!topologicalVertexStack.isEmpty()) { Vertex vertex = vertices.get(topologicalVertexStack.pop()); // infer credentials, resources and parallelism from data source Resource vertexTaskResource = vertex.getTaskResource(); if (vertexTaskResource == null) { vertexTaskResource = Resource.newInstance( tezConf.getInt(TezConfiguration.TEZ_TASK_RESOURCE_MEMORY_MB, TezConfiguration.TEZ_TASK_RESOURCE_MEMORY_MB_DEFAULT), tezConf.getInt(TezConfiguration.TEZ_TASK_RESOURCE_CPU_VCORES, TezConfiguration.TEZ_TASK_RESOURCE_CPU_VCORES_DEFAULT)); } Map<String, LocalResource> vertexLRs = Maps.newHashMap(); vertexLRs.putAll(vertex.getTaskLocalFiles()); List<DataSourceDescriptor> dataSources = vertex.getDataSources(); for (DataSourceDescriptor dataSource : dataSources) { if (dataSource.getCredentials() != null) { dagCredentials.addAll(dataSource.getCredentials()); } if (dataSource.getAdditionalLocalFiles() != null) { TezCommonUtils.addAdditionalLocalResources(dataSource.getAdditionalLocalFiles(), vertexLRs, "Vertex " + vertex.getName()); } } if (tezJarResources != null) { TezCommonUtils.addAdditionalLocalResources(tezJarResources, vertexLRs, "Vertex " + vertex.getName()); } if (binaryConfig != null) { vertexLRs.put(TezConstants.TEZ_PB_BINARY_CONF_NAME, binaryConfig); } int vertexParallelism = vertex.getParallelism(); VertexLocationHint vertexLocationHint = vertex.getLocationHint(); if (dataSources.size() == 1) { DataSourceDescriptor dataSource = dataSources.get(0); if (vertexParallelism == -1 && dataSource.getNumberOfShards() > -1) { vertexParallelism = dataSource.getNumberOfShards(); } if (vertexLocationHint == null && dataSource.getLocationHint() != null) { vertexLocationHint = dataSource.getLocationHint(); } } if (vertexParallelism == -1) { Preconditions.checkState(vertexLocationHint == null, "Cannot specify vertex location hint without specifying vertex parallelism. Vertex: " + vertex.getName()); } else if (vertexLocationHint != null) { Preconditions.checkState(vertexParallelism == vertexLocationHint.getTaskLocationHints().size(), "vertex task location hint must equal vertex parallelism. Vertex: " + vertex.getName()); } for (DataSinkDescriptor dataSink : vertex.getDataSinks()) { if (dataSink.getCredentials() != null) { dagCredentials.addAll(dataSink.getCredentials()); } } VertexPlan.Builder vertexBuilder = VertexPlan.newBuilder(); vertexBuilder.setName(vertex.getName()); vertexBuilder.setType(PlanVertexType.NORMAL); // vertex type is implicitly NORMAL until TEZ-46. vertexBuilder .setProcessorDescriptor(DagTypeConverters.convertToDAGPlan(vertex.getProcessorDescriptor())); if (vertex.getInputs().size() > 0) { for (RootInputLeafOutput<InputDescriptor, InputInitializerDescriptor> input : vertex.getInputs()) { vertexBuilder.addInputs(DagTypeConverters.convertToDAGPlan(input)); } } if (vertex.getOutputs().size() > 0) { for (RootInputLeafOutput<OutputDescriptor, OutputCommitterDescriptor> output : vertex .getOutputs()) { vertexBuilder.addOutputs(DagTypeConverters.convertToDAGPlan(output)); } } if (vertex.getConf() != null && vertex.getConf().size() > 0) { ConfigurationProto.Builder confBuilder = ConfigurationProto.newBuilder(); for (Map.Entry<String, String> entry : vertex.getConf().entrySet()) { PlanKeyValuePair.Builder keyValueBuilder = PlanKeyValuePair.newBuilder(); keyValueBuilder.setKey(entry.getKey()); keyValueBuilder.setValue(entry.getValue()); confBuilder.addConfKeyValues(keyValueBuilder); } vertexBuilder.setVertexConf(confBuilder); } //task config PlanTaskConfiguration.Builder taskConfigBuilder = PlanTaskConfiguration.newBuilder(); taskConfigBuilder.setNumTasks(vertexParallelism); taskConfigBuilder.setMemoryMb(vertexTaskResource.getMemory()); taskConfigBuilder.setVirtualCores(vertexTaskResource.getVirtualCores()); taskConfigBuilder.setJavaOpts( TezClientUtils.addDefaultsToTaskLaunchCmdOpts(vertex.getTaskLaunchCmdOpts(), tezConf)); taskConfigBuilder.setTaskModule(vertex.getName()); if (!vertexLRs.isEmpty()) { taskConfigBuilder.addAllLocalResource(DagTypeConverters.convertToDAGPlan(vertexLRs)); } Map<String, String> taskEnv = Maps.newHashMap(vertex.getTaskEnvironment()); TezYARNUtils.setupDefaultEnv(taskEnv, tezConf, TezConfiguration.TEZ_TASK_LAUNCH_ENV, TezConfiguration.TEZ_TASK_LAUNCH_ENV_DEFAULT, tezLrsAsArchive); for (Map.Entry<String, String> entry : taskEnv.entrySet()) { PlanKeyValuePair.Builder envSettingBuilder = PlanKeyValuePair.newBuilder(); envSettingBuilder.setKey(entry.getKey()); envSettingBuilder.setValue(entry.getValue()); taskConfigBuilder.addEnvironmentSetting(envSettingBuilder); } if (vertexLocationHint != null) { if (vertexLocationHint.getTaskLocationHints() != null) { for (TaskLocationHint hint : vertexLocationHint.getTaskLocationHints()) { PlanTaskLocationHint.Builder taskLocationHintBuilder = PlanTaskLocationHint.newBuilder(); // we can allow this later on if needed if (hint.getAffinitizedTask() != null) { throw new TezUncheckedException( "Task based affinity may not be specified via the DAG API"); } if (hint.getHosts() != null) { taskLocationHintBuilder.addAllHost(hint.getHosts()); } if (hint.getRacks() != null) { taskLocationHintBuilder.addAllRack(hint.getRacks()); } vertexBuilder.addTaskLocationHint(taskLocationHintBuilder); } } } if (vertex.getVertexManagerPlugin() != null) { vertexBuilder.setVertexManagerPlugin( DagTypeConverters.convertToDAGPlan(vertex.getVertexManagerPlugin())); } for (Edge inEdge : vertex.getInputEdges()) { vertexBuilder.addInEdgeId(inEdge.getId()); } for (Edge outEdge : vertex.getOutputEdges()) { vertexBuilder.addOutEdgeId(outEdge.getId()); } vertexBuilder.setTaskConfig(taskConfigBuilder); dagBuilder.addVertex(vertexBuilder); } for (Edge edge : edges) { EdgePlan.Builder edgeBuilder = EdgePlan.newBuilder(); edgeBuilder.setId(edge.getId()); edgeBuilder.setInputVertexName(edge.getInputVertex().getName()); edgeBuilder.setOutputVertexName(edge.getOutputVertex().getName()); edgeBuilder.setDataMovementType( DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getDataMovementType())); edgeBuilder.setDataSourceType( DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getDataSourceType())); edgeBuilder.setSchedulingType( DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getSchedulingType())); edgeBuilder.setEdgeSource(DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getEdgeSource())); edgeBuilder.setEdgeDestination( DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getEdgeDestination())); if (edge.getEdgeProperty().getDataMovementType() == DataMovementType.CUSTOM) { if (edge.getEdgeProperty().getEdgeManagerDescriptor() != null) { edgeBuilder.setEdgeManager( DagTypeConverters.convertToDAGPlan(edge.getEdgeProperty().getEdgeManagerDescriptor())); } // else the AM will deal with this. } dagBuilder.addEdge(edgeBuilder); } ConfigurationProto.Builder confProtoBuilder = ConfigurationProto.newBuilder(); if (dagAccessControls != null) { Configuration aclConf = new Configuration(false); dagAccessControls.serializeToConfiguration(aclConf); Iterator<Entry<String, String>> aclConfIter = aclConf.iterator(); while (aclConfIter.hasNext()) { Entry<String, String> entry = aclConfIter.next(); PlanKeyValuePair.Builder kvp = PlanKeyValuePair.newBuilder(); kvp.setKey(entry.getKey()); kvp.setValue(entry.getValue()); TezConfiguration.validateProperty(entry.getKey(), Scope.DAG); confProtoBuilder.addConfKeyValues(kvp); } } if (additionalConfigs != null && !additionalConfigs.isEmpty()) { for (Entry<String, String> entry : additionalConfigs.entrySet()) { PlanKeyValuePair.Builder kvp = PlanKeyValuePair.newBuilder(); kvp.setKey(entry.getKey()); kvp.setValue(entry.getValue()); TezConfiguration.validateProperty(entry.getKey(), Scope.DAG); confProtoBuilder.addConfKeyValues(kvp); } } if (this.dagConf != null && !this.dagConf.isEmpty()) { for (Entry<String, String> entry : this.dagConf.entrySet()) { PlanKeyValuePair.Builder kvp = PlanKeyValuePair.newBuilder(); kvp.setKey(entry.getKey()); kvp.setValue(entry.getValue()); confProtoBuilder.addConfKeyValues(kvp); } } dagBuilder.setDagConf(confProtoBuilder); if (dagCredentials != null) { dagBuilder.setCredentialsBinary(DagTypeConverters.convertCredentialsToProto(dagCredentials)); TezCommonUtils.logCredentials(LOG, dagCredentials, "dag"); } return dagBuilder.build(); }
From source file:org.apache.tez.dag.app.rm.container.AMContainerHelpers.java
License:Apache License
/** * Create the common {@link ContainerLaunchContext} for all attempts. * * @param applicationACLs/*from w w w. ja v a 2s .c o m*/ */ private static ContainerLaunchContext createCommonContainerLaunchContext( Map<ApplicationAccessType, String> applicationACLs, Credentials credentials, Map<String, LocalResource> localResources) { // Application environment Map<String, String> environment = new HashMap<String, String>(); // Service data Map<String, ByteBuffer> serviceData = new HashMap<String, ByteBuffer>(); // Tokens // Setup up task credentials buffer ByteBuffer containerCredentialsBuffer = ByteBuffer.wrap(new byte[] {}); try { Credentials containerCredentials = new Credentials(); // All Credentials need to be set so that YARN can localize the resources // correctly, even though they may not be used by all tasks which will run // on this container. LOG.info("Adding #" + credentials.numberOfTokens() + " tokens and #" + credentials.numberOfSecretKeys() + " secret keys for NM use for launching container"); containerCredentials.addAll(credentials); DataOutputBuffer containerTokens_dob = new DataOutputBuffer(); containerCredentials.writeTokenStorageToStream(containerTokens_dob); containerCredentialsBuffer = ByteBuffer.wrap(containerTokens_dob.getData(), 0, containerTokens_dob.getLength()); // Add shuffle token LOG.info("Putting shuffle token in serviceData"); serviceData.put(TezConstants.TEZ_SHUFFLE_HANDLER_SERVICE_ID, serializeServiceData(TokenCache.getSessionToken(containerCredentials))); } catch (IOException e) { throw new TezUncheckedException(e); } // Construct the actual Container // The null fields are per-container and will be constructed for each // container separately. ContainerLaunchContext container = ContainerLaunchContext.newInstance(localResources, environment, null, serviceData, containerCredentialsBuffer, applicationACLs); return container; }