Search in sources :

Example 1 with HdfsClient

use of org.apache.zeppelin.submarine.hadoop.HdfsClient in project zeppelin by apache.

the class HdfsClientTest method initEnv.

@BeforeClass
public static void initEnv() {
    ZeppelinConfiguration zconf = ZeppelinConfiguration.create();
    Properties properties = new Properties();
    hdfsClient = new HdfsClient(properties);
}
Also used : ZeppelinConfiguration(org.apache.zeppelin.conf.ZeppelinConfiguration) HdfsClient(org.apache.zeppelin.submarine.hadoop.HdfsClient) Properties(java.util.Properties) BeforeClass(org.junit.BeforeClass)

Example 2 with HdfsClient

use of org.apache.zeppelin.submarine.hadoop.HdfsClient in project zeppelin by apache.

the class JobRunThread method run.

public void run() {
    boolean tryLock = lockRunning.tryLock();
    if (false == tryLock) {
        LOGGER.warn("Can not get JobRunThread lockRunning!");
        return;
    }
    SubmarineUI submarineUI = submarineJob.getSubmarineUI();
    try {
        InterpreterContext intpContext = submarineJob.getIntpContext();
        String noteId = intpContext.getNoteId();
        String userName = intpContext.getAuthenticationInfo().getUser();
        String jobName = SubmarineUtils.getJobName(userName, noteId);
        if (true == running.get()) {
            String message = String.format("Job %s already running.", jobName);
            submarineUI.outputLog("WARN", message);
            LOGGER.warn(message);
            return;
        }
        running.set(true);
        Properties properties = submarineJob.getProperties();
        HdfsClient hdfsClient = submarineJob.getHdfsClient();
        File pythonWorkDir = submarineJob.getPythonWorkDir();
        submarineJob.setCurrentJobState(EXECUTE_SUBMARINE);
        String algorithmPath = properties.getProperty(SubmarineConstants.SUBMARINE_ALGORITHM_HDFS_PATH, "");
        if (!algorithmPath.startsWith("hdfs://")) {
            String message = "Algorithm file upload HDFS path, " + "Must be `hdfs://` prefix. now setting " + algorithmPath;
            submarineUI.outputLog("Configuration error", message);
            return;
        }
        List<ParagraphInfo> paragraphInfos = intpContext.getIntpEventClient().getParagraphList(userName, noteId);
        String outputMsg = hdfsClient.saveParagraphToFiles(noteId, paragraphInfos, pythonWorkDir == null ? "" : pythonWorkDir.getAbsolutePath(), properties);
        if (!StringUtils.isEmpty(outputMsg)) {
            submarineUI.outputLog("Save algorithm file", outputMsg);
        }
        HashMap jinjaParams = SubmarineUtils.propertiesToJinjaParams(properties, submarineJob, true);
        URL urlTemplate = Resources.getResource(SubmarineJob.SUBMARINE_JOBRUN_TF_JINJA);
        String template = Resources.toString(urlTemplate, Charsets.UTF_8);
        Jinjava jinjava = new Jinjava();
        String submarineCmd = jinjava.render(template, jinjaParams);
        // If the first line is a newline, delete the newline
        int firstLineIsNewline = submarineCmd.indexOf("\n");
        if (firstLineIsNewline == 0) {
            submarineCmd = submarineCmd.replaceFirst("\n", "");
        }
        StringBuffer sbLogs = new StringBuffer(submarineCmd);
        submarineUI.outputLog("Submarine submit command", sbLogs.toString());
        long timeout = Long.valueOf(properties.getProperty(SubmarineJob.TIMEOUT_PROPERTY, SubmarineJob.defaultTimeout));
        CommandLine cmdLine = CommandLine.parse(SubmarineJob.shell);
        cmdLine.addArgument(submarineCmd, false);
        DefaultExecutor executor = new DefaultExecutor();
        ExecuteWatchdog watchDog = new ExecuteWatchdog(timeout);
        executor.setWatchdog(watchDog);
        StringBuffer sbLogOutput = new StringBuffer();
        executor.setStreamHandler(new PumpStreamHandler(new LogOutputStream() {

            @Override
            protected void processLine(String line, int level) {
                line = line.trim();
                if (!StringUtils.isEmpty(line)) {
                    sbLogOutput.append(line + "\n");
                }
            }
        }));
        if (Boolean.valueOf(properties.getProperty(SubmarineJob.DIRECTORY_USER_HOME))) {
            executor.setWorkingDirectory(new File(System.getProperty("user.home")));
        }
        Map<String, String> env = new HashMap<>();
        String launchMode = (String) jinjaParams.get(SubmarineConstants.INTERPRETER_LAUNCH_MODE);
        if (StringUtils.equals(launchMode, "yarn")) {
            // Set environment variables in the submarine interpreter container run on yarn
            String javaHome, hadoopHome, hadoopConf;
            javaHome = (String) jinjaParams.get(SubmarineConstants.DOCKER_JAVA_HOME);
            hadoopHome = (String) jinjaParams.get(SubmarineConstants.DOCKER_HADOOP_HDFS_HOME);
            hadoopConf = (String) jinjaParams.get(SubmarineConstants.SUBMARINE_HADOOP_CONF_DIR);
            env.put("JAVA_HOME", javaHome);
            env.put("HADOOP_HOME", hadoopHome);
            env.put("HADOOP_HDFS_HOME", hadoopHome);
            env.put("HADOOP_CONF_DIR", hadoopConf);
            env.put("YARN_CONF_DIR", hadoopConf);
            env.put("CLASSPATH", "`$HADOOP_HDFS_HOME/bin/hadoop classpath --glob`");
            env.put("ZEPPELIN_FORCE_STOP", "true");
        }
        LOGGER.info("Execute EVN: {}, Command: {} ", env.toString(), submarineCmd);
        AtomicBoolean cmdLineRunning = new AtomicBoolean(true);
        executor.execute(cmdLine, env, new DefaultExecuteResultHandler() {

            @Override
            public void onProcessComplete(int exitValue) {
                String message = String.format("jobName %s ProcessComplete exit value is : %d", jobName, exitValue);
                LOGGER.info(message);
                submarineUI.outputLog("JOR RUN COMPLETE", message);
                cmdLineRunning.set(false);
                submarineJob.setCurrentJobState(EXECUTE_SUBMARINE_FINISHED);
            }

            @Override
            public void onProcessFailed(ExecuteException e) {
                String message = String.format("jobName %s ProcessFailed exit value is : %d, exception is : %s", jobName, e.getExitValue(), e.getMessage());
                LOGGER.error(message);
                submarineUI.outputLog("JOR RUN FAILED", message);
                cmdLineRunning.set(false);
                submarineJob.setCurrentJobState(EXECUTE_SUBMARINE_ERROR);
            }
        });
        int loopCount = 100;
        while ((loopCount-- > 0) && cmdLineRunning.get() && running.get()) {
            Thread.sleep(1000);
        }
        if (watchDog.isWatching()) {
            watchDog.destroyProcess();
            Thread.sleep(1000);
        }
        if (watchDog.isWatching()) {
            watchDog.killedProcess();
        }
        // Check if it has been submitted to YARN
        Map<String, Object> jobState = submarineJob.getJobStateByYarn(jobName);
        loopCount = 50;
        while ((loopCount-- > 0) && !jobState.containsKey("state") && running.get()) {
            Thread.sleep(3000);
            jobState = submarineJob.getJobStateByYarn(jobName);
        }
        if (!jobState.containsKey("state")) {
            String message = String.format("JOB %s was not submitted to YARN!", jobName);
            LOGGER.error(message);
            submarineUI.outputLog("JOR RUN FAILED", message);
            submarineJob.setCurrentJobState(EXECUTE_SUBMARINE_ERROR);
        }
    } catch (Exception e) {
        LOGGER.error(e.getMessage(), e);
        submarineJob.setCurrentJobState(EXECUTE_SUBMARINE_ERROR);
        submarineUI.outputLog("Exception", e.getMessage());
    } finally {
        running.set(false);
        lockRunning.unlock();
    }
}
Also used : HashMap(java.util.HashMap) Properties(java.util.Properties) URL(java.net.URL) PumpStreamHandler(org.apache.commons.exec.PumpStreamHandler) ExecuteException(org.apache.commons.exec.ExecuteException) InterpreterContext(org.apache.zeppelin.interpreter.InterpreterContext) DefaultExecuteResultHandler(org.apache.commons.exec.DefaultExecuteResultHandler) DefaultExecutor(org.apache.commons.exec.DefaultExecutor) ExecuteWatchdog(org.apache.commons.exec.ExecuteWatchdog) Jinjava(com.hubspot.jinjava.Jinjava) HdfsClient(org.apache.zeppelin.submarine.hadoop.HdfsClient) LogOutputStream(org.apache.commons.exec.LogOutputStream) ExecuteException(org.apache.commons.exec.ExecuteException) ParagraphInfo(org.apache.zeppelin.interpreter.thrift.ParagraphInfo) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) CommandLine(org.apache.commons.exec.CommandLine) File(java.io.File) SubmarineUI(org.apache.zeppelin.submarine.commons.SubmarineUI)

Aggregations

Properties (java.util.Properties)2 HdfsClient (org.apache.zeppelin.submarine.hadoop.HdfsClient)2 Jinjava (com.hubspot.jinjava.Jinjava)1 File (java.io.File)1 URL (java.net.URL)1 HashMap (java.util.HashMap)1 AtomicBoolean (java.util.concurrent.atomic.AtomicBoolean)1 CommandLine (org.apache.commons.exec.CommandLine)1 DefaultExecuteResultHandler (org.apache.commons.exec.DefaultExecuteResultHandler)1 DefaultExecutor (org.apache.commons.exec.DefaultExecutor)1 ExecuteException (org.apache.commons.exec.ExecuteException)1 ExecuteWatchdog (org.apache.commons.exec.ExecuteWatchdog)1 LogOutputStream (org.apache.commons.exec.LogOutputStream)1 PumpStreamHandler (org.apache.commons.exec.PumpStreamHandler)1 ZeppelinConfiguration (org.apache.zeppelin.conf.ZeppelinConfiguration)1 InterpreterContext (org.apache.zeppelin.interpreter.InterpreterContext)1 ParagraphInfo (org.apache.zeppelin.interpreter.thrift.ParagraphInfo)1 SubmarineUI (org.apache.zeppelin.submarine.commons.SubmarineUI)1 BeforeClass (org.junit.BeforeClass)1