Examples with StorageAgent com.datatorrent.api.StorageAgent used on opensource projects

Example 1 with StorageAgent

use of com.datatorrent.api.StorageAgent in project apex-core by apache.

the class StreamingContainerAgent method createOperatorDeployInfo.

/**
   * Create deploy info for operator.
   * <p>
   *
   * @return {@link com.datatorrent.stram.api.OperatorDeployInfo}
   */
private OperatorDeployInfo createOperatorDeployInfo(PTOperator oper) {
    OperatorDeployInfo ndi;
    if (oper.isUnifier()) {
        UnifierDeployInfo udi = new UnifierDeployInfo();
        /* the constructor auto sets the type */
        try {
            udi.operatorAttributes = oper.getUnifiedOperatorMeta().getAttributes().clone();
        } catch (CloneNotSupportedException ex) {
            throw new RuntimeException("Cannot clone unifier attributes", ex);
        }
        ndi = udi;
    } else {
        ndi = new OperatorDeployInfo();
        Operator operator = oper.getOperatorMeta().getOperator();
        if (operator instanceof InputOperator) {
            ndi.type = OperatorType.INPUT;
            if (!oper.getInputs().isEmpty()) {
                //we check if any input port is connected which would make it a Generic operator.
                for (PTOperator.PTInput ptInput : oper.getInputs()) {
                    if (ptInput.logicalStream != null && ptInput.logicalStream.getSource() != null) {
                        ndi.type = OperatorType.GENERIC;
                        break;
                    }
                }
            }
        } else {
            ndi.type = OperatorType.GENERIC;
        }
    }
    Checkpoint checkpoint = oper.getRecoveryCheckpoint();
    ProcessingMode pm = oper.getOperatorMeta().getValue(OperatorContext.PROCESSING_MODE);
    if (pm == ProcessingMode.AT_MOST_ONCE || pm == ProcessingMode.EXACTLY_ONCE) {
        // TODO: following should be handled in the container at deploy time
        // for exactly once container should also purge previous checkpoint
        // whenever new checkpoint is written.
        StorageAgent agent = oper.getOperatorMeta().getAttributes().get(OperatorContext.STORAGE_AGENT);
        if (agent == null) {
            agent = initCtx.getValue(OperatorContext.STORAGE_AGENT);
        }
        // pick checkpoint most recently written
        try {
            long[] windowIds = agent.getWindowIds(oper.getId());
            long checkpointId = Stateless.WINDOW_ID;
            for (long windowId : windowIds) {
                if (windowId > checkpointId) {
                    checkpointId = windowId;
                }
            }
            if (checkpoint == null || checkpoint.windowId != checkpointId) {
                checkpoint = new Checkpoint(checkpointId, 0, 0);
            }
        } catch (Exception e) {
            throw new RuntimeException("Failed to determine checkpoint window id " + oper, e);
        }
    }
    LOG.debug("{} recovery checkpoint {}", oper, checkpoint);
    ndi.checkpoint = checkpoint;
    ndi.name = oper.getOperatorMeta().getName();
    ndi.id = oper.getId();
    try {
        // clone map before modifying it
        ndi.contextAttributes = oper.getOperatorMeta().getAttributes().clone();
    } catch (CloneNotSupportedException ex) {
        throw new RuntimeException("Cannot clone operator attributes", ex);
    }
    if (oper.isOperatorStateLess()) {
        ndi.contextAttributes.put(OperatorContext.STATELESS, true);
    }
    return ndi;
}

Example 2 with StorageAgent

use of com.datatorrent.api.StorageAgent in project apex-core by apache.

the class Node method checkpoint.

void checkpoint(long windowId) {
    if (!context.stateless) {
        if (operator instanceof Operator.CheckpointNotificationListener) {
            ((Operator.CheckpointNotificationListener) operator).beforeCheckpoint(windowId);
        }
        StorageAgent ba = context.getValue(OperatorContext.STORAGE_AGENT);
        if (ba != null) {
            try {
                checkpointStats = new Stats.CheckpointStats();
                checkpointStats.checkpointStartTime = System.currentTimeMillis();
                ba.save(operator, id, windowId);
                if (ba instanceof AsyncStorageAgent) {
                    AsyncStorageAgent asyncStorageAgent = (AsyncStorageAgent) ba;
                    if (!asyncStorageAgent.isSyncCheckpoint()) {
                        if (PROCESSING_MODE != ProcessingMode.EXACTLY_ONCE) {
                            CheckpointWindowInfo checkpointWindowInfo = new CheckpointWindowInfo();
                            checkpointWindowInfo.windowId = windowId;
                            checkpointWindowInfo.applicationWindowCount = applicationWindowCount;
                            checkpointWindowInfo.checkpointWindowCount = checkpointWindowCount;
                            CheckpointHandler checkpointHandler = new CheckpointHandler();
                            checkpointHandler.agent = asyncStorageAgent;
                            checkpointHandler.operatorId = id;
                            checkpointHandler.windowId = windowId;
                            checkpointHandler.stats = checkpointStats;
                            FutureTask<Stats.CheckpointStats> futureTask = new FutureTask<>(checkpointHandler);
                            taskQueue.add(new Pair<>(futureTask, checkpointWindowInfo));
                            executorService.submit(futureTask);
                            checkpoint = null;
                            checkpointStats = null;
                            return;
                        } else {
                            asyncStorageAgent.flush(id, windowId);
                        }
                    }
                }
                checkpointStats.checkpointTime = System.currentTimeMillis() - checkpointStats.checkpointStartTime;
            } catch (IOException ie) {
                try {
                    logger.warn("Rolling back checkpoint {} for Operator {} due to the exception {}", Codec.getStringWindowId(windowId), operator, ie);
                    ba.delete(id, windowId);
                } catch (IOException ex) {
                    logger.warn("Error while rolling back checkpoint", ex);
                }
                throw new RuntimeException(ie);
            }
        }
    }
    calculateNextCheckpointWindow();
    dagCheckpointOffsetCount = 0;
    checkpoint = new Checkpoint(windowId, applicationWindowCount, checkpointWindowCount);
    if (operator instanceof Operator.CheckpointListener) {
        ((Operator.CheckpointListener) operator).checkpointed(windowId);
    }
}

Example 3 with StorageAgent

use of com.datatorrent.api.StorageAgent in project apex-core by apache.

the class StramRecoveryTest method checkpoint.

public static void checkpoint(StreamingContainerManager scm, PTOperator oper, Checkpoint checkpoint) throws Exception {
    // write checkpoint while AM is out,
    // it needs to be picked up as part of restore
    StorageAgent sa = oper.getOperatorMeta().getValue(OperatorContext.STORAGE_AGENT);
    sa.save(oper.getOperatorMeta().getOperator(), oper.getId(), checkpoint.windowId);
}

Example 4 with StorageAgent

use of com.datatorrent.api.StorageAgent in project apex-core by apache.

the class StreamingContainer method deployNodes.

private void deployNodes(List<OperatorDeployInfo> nodeList) throws IOException {
    for (OperatorDeployInfo ndi : nodeList) {
        StorageAgent backupAgent = getValue(OperatorContext.STORAGE_AGENT, ndi);
        assert (backupAgent != null);
        Context parentContext;
        if (ndi instanceof UnifierDeployInfo) {
            OperatorContext unifiedOperatorContext = new OperatorContext(0, ndi.name, ((UnifierDeployInfo) ndi).operatorAttributes, containerContext);
            parentContext = new PortContext(ndi.inputs.get(0).contextAttributes, unifiedOperatorContext);
            massageUnifierDeployInfo(ndi);
        } else {
            parentContext = containerContext;
        }
        OperatorContext ctx = new OperatorContext(ndi.id, ndi.name, ndi.contextAttributes, parentContext);
        ctx.attributes.put(OperatorContext.ACTIVATION_WINDOW_ID, ndi.checkpoint.windowId);
        logger.debug("Restoring operator {} to checkpoint {} stateless={}.", ndi.id, Codec.getStringWindowId(ndi.checkpoint.windowId), ctx.stateless);
        Node<?> node = Node.retrieveNode(backupAgent.load(ndi.id, ctx.stateless ? Stateless.WINDOW_ID : ndi.checkpoint.windowId), ctx, ndi.type);
        node.currentWindowId = ndi.checkpoint.windowId;
        node.applicationWindowCount = ndi.checkpoint.applicationWindowCount;
        node.firstWindowMillis = firstWindowMillis;
        node.windowWidthMillis = windowWidthMillis;
        node.setId(ndi.id);
        nodes.put(ndi.id, node);
        logger.debug("Marking operator {} as deployed.", node);
    }
}

Example 5 with StorageAgent

use of com.datatorrent.api.StorageAgent in project apex-core by apache.

the class StramClient method startApplication.

/**
   * Launch application for the dag represented by this client.
   *
   * @throws YarnException
   * @throws IOException
   */
public void startApplication() throws YarnException, IOException {
    Class<?>[] defaultClasses;
    if (applicationType.equals(YARN_APPLICATION_TYPE)) {
        //TODO restrict the security check to only check if security is enabled for webservices.
        if (UserGroupInformation.isSecurityEnabled()) {
            defaultClasses = APEX_SECURITY_CLASSES;
        } else {
            defaultClasses = APEX_CLASSES;
        }
    } else {
        throw new IllegalStateException(applicationType + " is not a valid application type.");
    }
    LinkedHashSet<String> localJarFiles = findJars(dag, defaultClasses);
    if (resources != null) {
        localJarFiles.addAll(resources);
    }
    YarnClusterMetrics clusterMetrics = yarnClient.getYarnClusterMetrics();
    LOG.info("Got Cluster metric info from ASM, numNodeManagers={}", clusterMetrics.getNumNodeManagers());
    //GetClusterNodesRequest clusterNodesReq = Records.newRecord(GetClusterNodesRequest.class);
    //GetClusterNodesResponse clusterNodesResp = rmClient.clientRM.getClusterNodes(clusterNodesReq);
    //LOG.info("Got Cluster node info from ASM");
    //for (NodeReport node : clusterNodesResp.getNodeReports()) {
    //  LOG.info("Got node report from ASM for"
    //           + ", nodeId=" + node.getNodeId()
    //           + ", nodeAddress" + node.getHttpAddress()
    //           + ", nodeRackName" + node.getRackName()
    //           + ", nodeNumContainers" + node.getNumContainers()
    //           + ", nodeHealthStatus" + node.getHealthReport());
    //}
    List<QueueUserACLInfo> listAclInfo = yarnClient.getQueueAclsInfo();
    for (QueueUserACLInfo aclInfo : listAclInfo) {
        for (QueueACL userAcl : aclInfo.getUserAcls()) {
            LOG.info("User ACL Info for Queue queueName={}, userAcl={}", aclInfo.getQueueName(), userAcl.name());
        }
    }
    // Get a new application id
    YarnClientApplication newApp = yarnClient.createApplication();
    appId = newApp.getNewApplicationResponse().getApplicationId();
    // Dump out information about cluster capability as seen by the resource manager
    int maxMem = newApp.getNewApplicationResponse().getMaximumResourceCapability().getMemory();
    LOG.info("Max mem capability of resources in this cluster " + maxMem);
    int amMemory = dag.getMasterMemoryMB();
    if (amMemory > maxMem) {
        LOG.info("AM memory specified above max threshold of cluster. Using max value, specified={}, max={}", amMemory, maxMem);
        amMemory = maxMem;
    }
    if (dag.getAttributes().get(LogicalPlan.APPLICATION_ID) == null) {
        dag.setAttribute(LogicalPlan.APPLICATION_ID, appId.toString());
    }
    // Create launch context for app master
    LOG.info("Setting up application submission context for ASM");
    ApplicationSubmissionContext appContext = Records.newRecord(ApplicationSubmissionContext.class);
    // set the application id
    appContext.setApplicationId(appId);
    // set the application name
    appContext.setApplicationName(dag.getValue(LogicalPlan.APPLICATION_NAME));
    appContext.setApplicationType(this.applicationType);
    if (YARN_APPLICATION_TYPE.equals(this.applicationType)) {
    //appContext.setMaxAppAttempts(1); // no retries until Stram is HA
    }
    appContext.setKeepContainersAcrossApplicationAttempts(true);
    // Set up the container launch context for the application master
    ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class);
    // application master launch.
    if (UserGroupInformation.isSecurityEnabled()) {
        Credentials credentials = new Credentials();
        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");
        }
        // For now, only getting tokens for the default file-system.
        try (FileSystem fs = StramClientUtils.newFileSystemInstance(conf)) {
            final Token<?>[] tokens = fs.addDelegationTokens(tokenRenewer, credentials);
            if (tokens != null) {
                for (Token<?> token : tokens) {
                    LOG.info("Got dt for " + fs.getUri() + "; " + token);
                }
            }
        }
        new ClientRMHelper(yarnClient, conf).addRMDelegationToken(tokenRenewer, credentials);
        DataOutputBuffer dob = new DataOutputBuffer();
        credentials.writeTokenStorageToStream(dob);
        ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
        amContainer.setTokens(fsTokens);
    }
    // Setup ACLs for the impersonating user
    LOG.debug("ACL login user {} current user {}", UserGroupInformation.getLoginUser(), UserGroupInformation.getCurrentUser());
    if (!UserGroupInformation.getCurrentUser().equals(UserGroupInformation.getLoginUser())) {
        ACLManager.setupUserACLs(amContainer, UserGroupInformation.getLoginUser().getShortUserName(), conf);
    }
    // set local resources for the application master
    // local files or archives as needed
    // In this scenario, the jar file for the application master is part of the local resources
    Map<String, LocalResource> localResources = new HashMap<>();
    // copy required jar files to dfs, to be localized for containers
    try (FileSystem fs = StramClientUtils.newFileSystemInstance(conf)) {
        Path appsBasePath = new Path(StramClientUtils.getDTDFSRootDir(fs, conf), StramClientUtils.SUBDIR_APPS);
        Path appPath;
        String configuredAppPath = dag.getValue(LogicalPlan.APPLICATION_PATH);
        if (configuredAppPath == null) {
            appPath = new Path(appsBasePath, appId.toString());
        } else {
            appPath = new Path(configuredAppPath);
        }
        String libJarsCsv = copyFromLocal(fs, appPath, localJarFiles.toArray(new String[] {}));
        LOG.info("libjars: {}", libJarsCsv);
        dag.getAttributes().put(Context.DAGContext.LIBRARY_JARS, libJarsCsv);
        LaunchContainerRunnable.addFilesToLocalResources(LocalResourceType.FILE, libJarsCsv, localResources, fs);
        if (archives != null) {
            String[] localFiles = archives.split(",");
            String archivesCsv = copyFromLocal(fs, appPath, localFiles);
            LOG.info("archives: {}", archivesCsv);
            dag.getAttributes().put(LogicalPlan.ARCHIVES, archivesCsv);
            LaunchContainerRunnable.addFilesToLocalResources(LocalResourceType.ARCHIVE, archivesCsv, localResources, fs);
        }
        if (files != null) {
            String[] localFiles = files.split(",");
            String filesCsv = copyFromLocal(fs, appPath, localFiles);
            LOG.info("files: {}", filesCsv);
            dag.getAttributes().put(LogicalPlan.FILES, filesCsv);
            LaunchContainerRunnable.addFilesToLocalResources(LocalResourceType.FILE, filesCsv, localResources, fs);
        }
        dag.getAttributes().put(LogicalPlan.APPLICATION_PATH, appPath.toString());
        StorageAgent agent = dag.getAttributes().get(OperatorContext.STORAGE_AGENT);
        if (agent != null && agent instanceof StorageAgent.ApplicationAwareStorageAgent) {
            ((StorageAgent.ApplicationAwareStorageAgent) agent).setApplicationAttributes(dag.getAttributes());
        }
        if (dag.getAttributes().get(OperatorContext.STORAGE_AGENT) == null) {
            /* which would be the most likely case */
            Path checkpointPath = new Path(appPath, LogicalPlan.SUBDIR_CHECKPOINTS);
            // use conf client side to pickup any proxy settings from dt-site.xml
            dag.setAttribute(OperatorContext.STORAGE_AGENT, new AsyncFSStorageAgent(checkpointPath.toString(), conf));
        }
        if (dag.getAttributes().get(LogicalPlan.CONTAINER_OPTS_CONFIGURATOR) == null) {
            dag.setAttribute(LogicalPlan.CONTAINER_OPTS_CONFIGURATOR, new BasicContainerOptConfigurator());
        }
        // Set the log4j properties if needed
        if (!log4jPropFile.isEmpty()) {
            Path log4jSrc = new Path(log4jPropFile);
            Path log4jDst = new Path(appPath, "log4j.props");
            fs.copyFromLocalFile(false, true, log4jSrc, log4jDst);
            FileStatus log4jFileStatus = fs.getFileStatus(log4jDst);
            LocalResource log4jRsrc = Records.newRecord(LocalResource.class);
            log4jRsrc.setType(LocalResourceType.FILE);
            log4jRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
            log4jRsrc.setResource(ConverterUtils.getYarnUrlFromURI(log4jDst.toUri()));
            log4jRsrc.setTimestamp(log4jFileStatus.getModificationTime());
            log4jRsrc.setSize(log4jFileStatus.getLen());
            localResources.put("log4j.properties", log4jRsrc);
        }
        if (originalAppId != null) {
            Path origAppPath = new Path(appsBasePath, this.originalAppId);
            LOG.info("Restart from {}", origAppPath);
            copyInitialState(origAppPath);
        }
        // push logical plan to DFS location
        Path cfgDst = new Path(appPath, LogicalPlan.SER_FILE_NAME);
        FSDataOutputStream outStream = fs.create(cfgDst, true);
        LogicalPlan.write(this.dag, outStream);
        outStream.close();
        Path launchConfigDst = new Path(appPath, LogicalPlan.LAUNCH_CONFIG_FILE_NAME);
        outStream = fs.create(launchConfigDst, true);
        conf.writeXml(outStream);
        outStream.close();
        LaunchContainerRunnable.addFileToLocalResources(LogicalPlan.SER_FILE_NAME, fs.getFileStatus(cfgDst), LocalResourceType.FILE, localResources);
        // Set local resource info into app master container launch context
        amContainer.setLocalResources(localResources);
        // Set the necessary security tokens as needed
        //amContainer.setContainerTokens(containerToken);
        // Set the env variables to be setup in the env where the application master will be run
        LOG.info("Set the environment for the application master");
        Map<String, String> env = new HashMap<>();
        // Add application jar(s) location to classpath
        // At some point we should not be required to add
        // the hadoop specific classpaths to the env.
        // It should be provided out of the box.
        // For now setting all required classpaths including
        // the classpath to "." for the application jar(s)
        // including ${CLASSPATH} will duplicate the class path in app master, removing it for now
        //StringBuilder classPathEnv = new StringBuilder("${CLASSPATH}:./*");
        StringBuilder classPathEnv = new StringBuilder("./*");
        String classpath = conf.get(YarnConfiguration.YARN_APPLICATION_CLASSPATH);
        for (String c : StringUtils.isBlank(classpath) ? YarnConfiguration.DEFAULT_YARN_APPLICATION_CLASSPATH : classpath.split(",")) {
            if (c.equals("$HADOOP_CLIENT_CONF_DIR")) {
                // SPOI-2501
                continue;
            }
            classPathEnv.append(':');
            classPathEnv.append(c.trim());
        }
        env.put("CLASSPATH", classPathEnv.toString());
        // propagate to replace node managers user name (effective in non-secure mode)
        // also to indicate original login user during impersonation and important for setting ACLs
        env.put("HADOOP_USER_NAME", UserGroupInformation.getLoginUser().getUserName());
        amContainer.setEnvironment(env);
        // Set the necessary command to execute the application master
        ArrayList<CharSequence> vargs = new ArrayList<>(30);
        // Set java executable command
        LOG.info("Setting up app master command");
        vargs.add(javaCmd);
        if (dag.isDebug()) {
            vargs.add("-agentlib:jdwp=transport=dt_socket,server=y,suspend=n");
        }
        // default heap size 75% of total memory
        if (dag.getMasterJVMOptions() != null) {
            vargs.add(dag.getMasterJVMOptions());
        }
        Path tmpDir = new Path(ApplicationConstants.Environment.PWD.$(), YarnConfiguration.DEFAULT_CONTAINER_TEMP_DIR);
        vargs.add("-Djava.io.tmpdir=" + tmpDir);
        vargs.add("-Xmx" + (amMemory * 3 / 4) + "m");
        vargs.add("-XX:+HeapDumpOnOutOfMemoryError");
        vargs.add("-XX:HeapDumpPath=" + System.getProperty("java.io.tmpdir") + "/dt-heap-" + appId.getId() + ".bin");
        vargs.add("-Dhadoop.root.logger=" + (dag.isDebug() ? "DEBUG" : "INFO") + ",RFA");
        vargs.add("-Dhadoop.log.dir=" + ApplicationConstants.LOG_DIR_EXPANSION_VAR);
        vargs.add(String.format("-D%s=%s", StreamingContainer.PROP_APP_PATH, dag.assertAppPath()));
        StramClientUtils.addAttributeToArgs(LogicalPlan.APPLICATION_NAME, dag, vargs);
        StramClientUtils.addAttributeToArgs(LogicalPlan.LOGGER_APPENDER, dag, vargs);
        if (dag.isDebug()) {
            vargs.add("-Dlog4j.debug=true");
        }
        String loggersLevel = conf.get(StramUtils.DT_LOGGERS_LEVEL);
        if (loggersLevel != null) {
            vargs.add(String.format("-D%s=%s", StramUtils.DT_LOGGERS_LEVEL, loggersLevel));
        }
        vargs.add(StreamingAppMaster.class.getName());
        vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stdout");
        vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stderr");
        // Get final command
        StringBuilder command = new StringBuilder(9 * vargs.size());
        for (CharSequence str : vargs) {
            command.append(str).append(" ");
        }
        LOG.info("Completed setting up app master command " + command.toString());
        List<String> commands = new ArrayList<>();
        commands.add(command.toString());
        amContainer.setCommands(commands);
        // Set up resource type requirements
        // For now, only memory is supported so we set memory requirements
        Resource capability = Records.newRecord(Resource.class);
        capability.setMemory(amMemory);
        appContext.setResource(capability);
        // Service data is a binary blob that can be passed to the application
        // Not needed in this scenario
        // amContainer.setServiceData(serviceData);
        appContext.setAMContainerSpec(amContainer);
        // Set the priority for the application master
        Priority pri = Records.newRecord(Priority.class);
        pri.setPriority(amPriority);
        appContext.setPriority(pri);
        // Set the queue to which this application is to be submitted in the RM
        appContext.setQueue(queueName);
        // set the application tags
        appContext.setApplicationTags(tags);
        // Submit the application to the applications manager
        // SubmitApplicationResponse submitResp = rmClient.submitApplication(appRequest);
        // Ignore the response as either a valid response object is returned on success
        // or an exception thrown to denote some form of a failure
        String specStr = Objects.toStringHelper("Submitting application: ").add("name", appContext.getApplicationName()).add("queue", appContext.getQueue()).add("user", UserGroupInformation.getLoginUser()).add("resource", appContext.getResource()).toString();
        LOG.info(specStr);
        if (dag.isDebug()) {
        //LOG.info("Full submission context: " + appContext);
        }
        yarnClient.submitApplication(appContext);
    }
}