public static void main(String[] args) throws IOException {
Path baseDir = null;
String localPath = null;
String preservePath = null;
String sIgnoreTablesFilename = null;
String sNoPreserveFilename = null;
String sDateString = null;
long size = 0;
// UNIX dates for right now
long now = new java.util.Date().getTime() / 1000;
long maxDate = now;
for (int i = 0; i < args.length; i++) {
if (args[i].equals("--hdfs-path")) {
baseDir = new Path(args[++i]);
continue;
}
if (args[i].equals("--local-path")) {
localPath = args[++i];
continue;
}
if (args[i].equals("--preserve-path")) {
preservePath = args[++i];
continue;
}
if (args[i].equals("--no-preserve")) {
sNoPreserveFilename = args[++i];
continue;
}
if (args[i].equals("--ignore-tables")) {
sIgnoreTablesFilename = args[++i];
continue;
}
if (args[i].equals("--sleep")) {
try {
m_nSleepSeconds = Integer.parseInt(args[++i]);
} catch (Exception e) {
System.err.println("ERROR: " + e.toString() + "\n");
usage();
}
continue;
}
if (args[i].equals("--dry-run")) {
m_bDryRun = true;
continue;
}
if (args[i].equals("--date")) {
sDateString = args[++i];
continue;
}
if (args[i].equals("--max-date")) {
maxDate = Long.parseLong(args[++i]);
continue;
}
if (args[i].equals("--max-bytes")) {
size = Long.parseLong(args[++i]);
continue;
}
System.err.println("ERROR: unknown arg " + args[i]);
usage();
}
if (baseDir == null || localPath == null || preservePath == null || sDateString == null) {
usage();
}
long minDate;
if ("yesterday".equals(sDateString)) {
// figure out yesterday's dates
Calendar cal = Calendar.getInstance();
cal.roll(Calendar.DAY_OF_YEAR, -1);
// yesterday midnight
cal.set(Calendar.HOUR_OF_DAY, 0);
cal.set(Calendar.MINUTE, 0);
cal.set(Calendar.SECOND, 0);
cal.set(Calendar.MILLISECOND, 0);
minDate = cal.getTimeInMillis() / 1000;
// yesterday end of day
cal.set(Calendar.HOUR_OF_DAY, 23);
cal.set(Calendar.MINUTE, 59);
cal.set(Calendar.SECOND, 59);
cal.set(Calendar.MILLISECOND, 999);
maxDate = cal.getTimeInMillis() / 1000;
} else if ("last-week".equals(sDateString)) {
minDate = maxDate - (7 * 24 * 60 * 60);
} else if ("last-day".equals(sDateString)) {
minDate = maxDate - (24 * 60 * 60);
} else {
// UNIX date since epoch of last backup
minDate = Long.parseLong(sDateString);
}
long tmpDate = 0;
BackupHdfs bak = new BackupHdfs();
// initialize the list of tables to ignore
if (sIgnoreTablesFilename != null) {
bak.initializeTablesToIgnore(sIgnoreTablesFilename);
}
// initialize list of files to not preserve
if (sNoPreserveFilename != null) {
bak.initializeNoPreserve(sNoPreserveFilename);
}
ArrayList<Path> pathList = new ArrayList<Path>(2000);
HashMap<Path, Long> hmTimestamps = new HashMap<Path, Long>();
Configuration conf = new Configuration();
FileSystem fs = FileSystem.get(conf);
// If the HDFS path is a dir continue
if (fs.getFileStatus(baseDir).isDir()) {
Calendar cal = Calendar.getInstance();
System.err.println("");
cal.setTimeInMillis(minDate * 1000);
System.err.println("min date = " + cal.getTime().toString());
cal.setTimeInMillis(maxDate * 1000);
System.err.println("max date = " + cal.getTime().toString());
System.err.println("");
System.err.println("Searching filesystem: " + baseDir.toUri().getPath());
bak.checkDir(fs, minDate, maxDate, baseDir, pathList, hmTimestamps);
System.err.println("");
System.err.println("Skipped " + m_nIgnoredTables + " files due to ignored tables");
System.err.println("");
System.err.println("Number of files to backup = " + pathList.size());
System.err.println("Total bytes to backup = " + prettyPrintBytes(m_nTotalBytes));
System.err.println("");
System.err.println("sorting list of files...");
Collections.sort(pathList, new DateComparator(hmTimestamps));
System.err.println("done");
System.err.println("");
System.err.println("starting backup...");
tmpDate = bak.backupFiles(localPath, preservePath, fs, pathList, size);
bak.closeFiles();
System.err.println("");
System.err.println("backup completed...");
}
if (tmpDate == 0) {
// If not size limit reached print out date for right now
System.out.println(maxDate);
} else {
// Print out date for last file backed up
System.err.println("Size limit reached.");
System.out.println(tmpDate);
}
System.exit(0);
}
/** check a candidate plugin for jar hell before installing it */
private void jarHellCheck(Path candidate, boolean isolated) throws IOException {
// create list of current jars in classpath
final List<URL> jars = new ArrayList<>(Arrays.asList(JarHell.parseClassPath()));
// read existing bundles. this does some checks on the installation too.
List<Bundle> bundles = PluginsService.getPluginBundles(environment.pluginsFile());
// if we aren't isolated, we need to jarhellcheck against any other non-isolated plugins
// thats always the first bundle
if (isolated == false) {
jars.addAll(bundles.get(0).urls);
}
// add plugin jars to the list
Path pluginJars[] = FileSystemUtils.files(candidate, "*.jar");
for (Path jar : pluginJars) {
jars.add(jar.toUri().toURL());
}
// check combined (current classpath + new jars to-be-added)
try {
JarHell.checkJarHell(jars.toArray(new URL[jars.size()]));
} catch (Exception ex) {
throw new RuntimeException(ex);
}
}
/**
* Add a file path to the current set of classpath entries. It adds the file to cache as well.
* Intended to be used by user code.
*
* @param file Path of the file to be added
* @param conf Configuration that contains the classpath setting
* @param fs FileSystem with respect to which {@code archivefile} should be interpreted.
*/
public static void addFileToClassPath(Path file, Configuration conf, FileSystem fs)
throws IOException {
String filepath = file.toUri().getPath();
String classpath = conf.get("mapred.job.classpath.files");
conf.set(
"mapred.job.classpath.files",
classpath == null ? filepath : classpath + System.getProperty("path.separator") + filepath);
URI uri = fs.makeQualified(file).toUri();
addCacheFile(uri, conf);
}
/**
* Make a path relative with respect to a root path. absPath is always assumed to descend from
* root. Otherwise returned path is null.
*/
static String makeRelative(Path root, Path absPath) {
if (!absPath.isAbsolute()) {
throw new IllegalArgumentException("!absPath.isAbsolute(), absPath=" + absPath);
}
String p = absPath.toUri().getPath();
StringTokenizer pathTokens = new StringTokenizer(p, "/");
for (StringTokenizer rootTokens = new StringTokenizer(root.toUri().getPath(), "/");
rootTokens.hasMoreTokens(); ) {
if (!rootTokens.nextToken().equals(pathTokens.nextToken())) {
return null;
}
}
StringBuilder sb = new StringBuilder();
for (; pathTokens.hasMoreTokens(); ) {
sb.append(pathTokens.nextToken());
if (pathTokens.hasMoreTokens()) {
sb.append(Path.SEPARATOR);
}
}
return sb.length() == 0 ? "." : sb.toString();
}
/* this makes a path qualified in the har filesystem
* (non-Javadoc)
* @see org.apache.hadoop.fs.FilterFileSystem#makeQualified(
* org.apache.hadoop.fs.Path)
*/
@Override
public Path makeQualified(Path path) {
// make sure that we just get the
// path component
Path fsPath = path;
if (!path.isAbsolute()) {
fsPath = new Path(archivePath, path);
}
URI tmpURI = fsPath.toUri();
// change this to Har uri
return new Path(uri.getScheme(), harAuth, tmpURI.getPath());
}
/**
* Add an archive path to the current set of classpath entries. It adds the archive to cache as
* well. Intended to be used by user code.
*
* @param archive Path of the archive to be added
* @param conf Configuration that contains the classpath setting
* @param fs FileSystem with respect to which {@code archive} should be interpreted.
*/
public static void addArchiveToClassPath(Path archive, Configuration conf, FileSystem fs)
throws IOException {
String archivepath = archive.toUri().getPath();
String classpath = conf.get("mapred.job.classpath.archives");
conf.set(
"mapred.job.classpath.archives",
classpath == null
? archivepath
: classpath + System.getProperty("path.separator") + archivepath);
URI uri = fs.makeQualified(archive).toUri();
addCacheArchive(uri, conf);
}
private static URI addArchiveToClassPathHelper(Path archive, Configuration conf)
throws IOException {
String classpath = conf.get("mapred.job.classpath.archives");
// the scheme/authority use ':' as separator. put the unqualified path in classpath
String archivePath = archive.toUri().getPath();
conf.set(
"mapred.job.classpath.archives",
classpath == null
? archivePath
: classpath + System.getProperty("path.separator") + archivePath);
return archive.makeQualified(archive.getFileSystem(conf)).toUri();
}
/**
* this method returns the path inside the har filesystem. this is relative path inside the har
* filesystem.
*
* @param path the fully qualified path in the har filesystem.
* @return relative path in the filesystem.
*/
private Path getPathInHar(Path path) {
Path harPath = new Path(path.toUri().getPath());
if (archivePath.compareTo(harPath) == 0) return new Path(Path.SEPARATOR);
Path tmp = new Path(harPath.getName());
Path parent = harPath.getParent();
while (!(parent.compareTo(archivePath) == 0)) {
if (parent.toString().equals(Path.SEPARATOR)) {
tmp = null;
break;
}
tmp = new Path(parent.getName(), tmp);
parent = parent.getParent();
}
if (tmp != null) tmp = new Path(Path.SEPARATOR, tmp);
return tmp;
}
/**
* Initialize a Har filesystem per har archive. The archive home directory is the top level
* directory in the filesystem that contains the HAR archive. Be careful with this method, you do
* not want to go on creating new Filesystem instances per call to path.getFileSystem(). the uri
* of Har is har://underlyingfsscheme-host:port/archivepath. or har:///archivepath. This assumes
* the underlying filesystem to be used in case not specified.
*/
@Override
public void initialize(URI name, Configuration conf) throws IOException {
// initialize the metadata cache, if needed
initializeMetadataCache(conf);
// decode the name
URI underLyingURI = decodeHarURI(name, conf);
// we got the right har Path- now check if this is
// truly a har filesystem
Path harPath = archivePath(new Path(name.getScheme(), name.getAuthority(), name.getPath()));
if (harPath == null) {
throw new IOException("Invalid path for the Har Filesystem. " + name.toString());
}
if (fs == null) {
fs = FileSystem.get(underLyingURI, conf);
}
uri = harPath.toUri();
archivePath = new Path(uri.getPath());
harAuth = getHarAuth(underLyingURI);
// check for the underlying fs containing
// the index file
Path masterIndexPath = new Path(archivePath, "_masterindex");
Path archiveIndexPath = new Path(archivePath, "_index");
if (!fs.exists(masterIndexPath) || !fs.exists(archiveIndexPath)) {
throw new IOException(
"Invalid path for the Har Filesystem. " + "No index file in " + harPath);
}
metadata = harMetaCache.get(uri);
if (metadata != null) {
FileStatus mStat = fs.getFileStatus(masterIndexPath);
FileStatus aStat = fs.getFileStatus(archiveIndexPath);
if (mStat.getModificationTime() != metadata.getMasterIndexTimestamp()
|| aStat.getModificationTime() != metadata.getArchiveIndexTimestamp()) {
// the archive has been overwritten since we last read it
// remove the entry from the meta data cache
metadata = null;
harMetaCache.remove(uri);
}
}
if (metadata == null) {
metadata = new HarMetaData(fs, masterIndexPath, archiveIndexPath);
metadata.parseMetaData();
harMetaCache.put(uri, metadata);
}
}
public void copyInitialState(Path origAppDir) throws IOException {
// locate previous snapshot
String newAppDir = this.dag.assertAppPath();
FSRecoveryHandler recoveryHandler = new FSRecoveryHandler(origAppDir.toString(), conf);
// read snapshot against new dependencies
Object snapshot = recoveryHandler.restore();
if (snapshot == null) {
throw new IllegalArgumentException("No previous application state found in " + origAppDir);
}
InputStream logIs = recoveryHandler.getLog();
// modify snapshot state to switch app id
((StreamingContainerManager.CheckpointState) snapshot).setApplicationId(this.dag, conf);
Path checkpointPath = new Path(newAppDir, LogicalPlan.SUBDIR_CHECKPOINTS);
FileSystem fs = FileSystem.newInstance(origAppDir.toUri(), conf);
// remove the path that was created by the storage agent during deserialization and replacement
fs.delete(checkpointPath, true);
// write snapshot to new location
recoveryHandler = new FSRecoveryHandler(newAppDir, conf);
recoveryHandler.save(snapshot);
OutputStream logOs = recoveryHandler.rotateLog();
IOUtils.copy(logIs, logOs);
logOs.flush();
logOs.close();
logIs.close();
// copy sub directories that are not present in target
FileStatus[] lFiles = fs.listStatus(origAppDir);
for (FileStatus f : lFiles) {
if (f.isDirectory()) {
String targetPath = f.getPath().toString().replace(origAppDir.toString(), newAppDir);
if (!fs.exists(new Path(targetPath))) {
LOG.debug("Copying {} to {}", f.getPath(), targetPath);
FileUtil.copy(fs, f.getPath(), fs, new Path(targetPath), false, conf);
// FSUtil.copy(fs, f, fs, new Path(targetPath), false, false, conf);
} else {
LOG.debug("Ignoring {} as it already exists under {}", f.getPath(), targetPath);
// FSUtil.setPermission(fs, new Path(targetPath), new FsPermission((short)0777));
}
}
}
}
/**
* Delete a local path with asyncDiskService if available, or otherwise synchronously with local
* file system.
*/
private static void deleteLocalPath(
MRAsyncDiskService asyncDiskService, LocalFileSystem fs, Path path) throws IOException {
boolean deleted = false;
if (asyncDiskService != null) {
// Try to delete using asyncDiskService
String localPathToDelete = path.toUri().getPath();
deleted = asyncDiskService.moveAndDeleteAbsolutePath(localPathToDelete);
if (!deleted) {
LOG.warn(
"Cannot find DistributedCache path "
+ localPathToDelete
+ " on any of the asyncDiskService volumes!");
}
}
if (!deleted) {
// If no asyncDiskService, we will delete the files synchronously
fs.delete(path, true);
}
LOG.info("Deleted path " + path);
}
/**
* 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 = DATATORRENT_SECURITY_CLASSES;
} else {
defaultClasses = DATATORRENT_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="
+ aclInfo.getQueueName()
+ ", userAcl="
+ 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 capabililty 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="
+ amMemory
+ ", max="
+ 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
}
// Set up the container launch context for the application master
ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class);
// Setup security tokens
// If security is enabled get ResourceManager and NameNode delegation tokens.
// Set these tokens on the container so that they are sent as part of application submission.
// This also sets them up for renewal by ResourceManager. The NameNode delegation rmToken
// is also used by ResourceManager to fetch the jars from HDFS and set them up for the
// 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.
FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
try {
final Token<?> tokens[] = fs.addDelegationTokens(tokenRenewer, credentials);
if (tokens != null) {
for (Token<?> token : tokens) {
LOG.info("Got dt for " + fs.getUri() + "; " + token);
}
}
} finally {
fs.close();
}
addRMDelegationToken(tokenRenewer, credentials);
DataOutputBuffer dob = new DataOutputBuffer();
credentials.writeTokenStorageToStream(dob);
ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
amContainer.setTokens(fsTokens);
}
// 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<String, LocalResource>();
// copy required jar files to dfs, to be localized for containers
FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
try {
Path appsBasePath =
new Path(StramClientUtils.getDTDFSRootDir(fs, conf), StramClientUtils.SUBDIR_APPS);
Path appPath = new Path(appsBasePath, appId.toString());
String libJarsCsv = copyFromLocal(fs, appPath, localJarFiles.toArray(new String[] {}));
LOG.info("libjars: {}", libJarsCsv);
dag.getAttributes().put(LogicalPlan.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());
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 FSStorageAgent(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();
FileStatus topologyFileStatus = fs.getFileStatus(cfgDst);
LocalResource topologyRsrc = Records.newRecord(LocalResource.class);
topologyRsrc.setType(LocalResourceType.FILE);
topologyRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
topologyRsrc.setResource(ConverterUtils.getYarnUrlFromURI(cfgDst.toUri()));
topologyRsrc.setTimestamp(topologyFileStatus.getModificationTime());
topologyRsrc.setSize(topologyFileStatus.getLen());
localResources.put(LogicalPlan.SER_FILE_NAME, topologyRsrc);
// 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<String, String>();
// 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)
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<CharSequence>(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");
}
// Set Xmx based on am memory size
// default heap size 75% of total memory
vargs.add("-Xmx" + (amMemory * 3 / 4) + "m");
vargs.add("-XX:+HeapDumpOnOutOfMemoryError");
vargs.add("-XX:HeapDumpPath=/tmp/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()));
if (dag.isDebug()) {
vargs.add("-Dlog4j.debug=true");
}
String loggersLevel = conf.get(DTLoggerFactory.DT_LOGGERS_LEVEL);
if (loggersLevel != null) {
vargs.add(String.format("-D%s=%s", DTLoggerFactory.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<String>();
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);
// 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);
} finally {
fs.close();
}
}
/**
* Initialize DFSCopyFileMapper specific job-configuration.
*
* @param conf : The dfs/mapred configuration.
* @param jobConf : The handle to the jobConf object to be initialized.
* @param args Arguments
*/
private static void setup(Configuration conf, JobConf jobConf, final Arguments args)
throws IOException {
jobConf.set(DST_DIR_LABEL, args.dst.toUri().toString());
// set boolean values
final boolean update = args.flags.contains(Options.UPDATE);
final boolean overwrite = !update && args.flags.contains(Options.OVERWRITE);
jobConf.setBoolean(Options.UPDATE.propertyname, update);
jobConf.setBoolean(Options.OVERWRITE.propertyname, overwrite);
jobConf.setBoolean(
Options.IGNORE_READ_FAILURES.propertyname,
args.flags.contains(Options.IGNORE_READ_FAILURES));
jobConf.setBoolean(
Options.PRESERVE_STATUS.propertyname, args.flags.contains(Options.PRESERVE_STATUS));
final String randomId = getRandomId();
JobClient jClient = new JobClient(jobConf);
Path jobDirectory = new Path(jClient.getSystemDir(), NAME + "_" + randomId);
jobConf.set(JOB_DIR_LABEL, jobDirectory.toString());
FileSystem dstfs = args.dst.getFileSystem(conf);
boolean dstExists = dstfs.exists(args.dst);
boolean dstIsDir = false;
if (dstExists) {
dstIsDir = dstfs.getFileStatus(args.dst).isDir();
}
// default logPath
Path logPath = args.log;
if (logPath == null) {
String filename = "_distcp_logs_" + randomId;
if (!dstExists || !dstIsDir) {
Path parent = args.dst.getParent();
if (!dstfs.exists(parent)) {
dstfs.mkdirs(parent);
}
logPath = new Path(parent, filename);
} else {
logPath = new Path(args.dst, filename);
}
}
FileOutputFormat.setOutputPath(jobConf, logPath);
// create src list, dst list
FileSystem jobfs = jobDirectory.getFileSystem(jobConf);
Path srcfilelist = new Path(jobDirectory, "_distcp_src_files");
jobConf.set(SRC_LIST_LABEL, srcfilelist.toString());
SequenceFile.Writer src_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
srcfilelist,
LongWritable.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
Path dstfilelist = new Path(jobDirectory, "_distcp_dst_files");
SequenceFile.Writer dst_writer =
SequenceFile.createWriter(
jobfs, jobConf, dstfilelist, Text.class, Text.class, SequenceFile.CompressionType.NONE);
Path dstdirlist = new Path(jobDirectory, "_distcp_dst_dirs");
jobConf.set(DST_DIR_LIST_LABEL, dstdirlist.toString());
SequenceFile.Writer dir_writer =
SequenceFile.createWriter(
jobfs,
jobConf,
dstdirlist,
Text.class,
FilePair.class,
SequenceFile.CompressionType.NONE);
// handle the case where the destination directory doesn't exist
// and we've only a single src directory OR we're updating/overwriting
// the contents of the destination directory.
final boolean special = (args.srcs.size() == 1 && !dstExists) || update || overwrite;
int srcCount = 0, cnsyncf = 0, dirsyn = 0;
long fileCount = 0L, byteCount = 0L, cbsyncs = 0L;
try {
for (Iterator<Path> srcItr = args.srcs.iterator(); srcItr.hasNext(); ) {
final Path src = srcItr.next();
FileSystem srcfs = src.getFileSystem(conf);
FileStatus srcfilestat = srcfs.getFileStatus(src);
Path root = special && srcfilestat.isDir() ? src : src.getParent();
if (srcfilestat.isDir()) {
++srcCount;
}
Stack<FileStatus> pathstack = new Stack<FileStatus>();
for (pathstack.push(srcfilestat); !pathstack.empty(); ) {
FileStatus cur = pathstack.pop();
FileStatus[] children = srcfs.listStatus(cur.getPath());
for (int i = 0; i < children.length; i++) {
boolean skipfile = false;
final FileStatus child = children[i];
final String dst = makeRelative(root, child.getPath());
++srcCount;
if (child.isDir()) {
pathstack.push(child);
} else {
// skip file if the src and the dst files are the same.
skipfile = update && sameFile(srcfs, child, dstfs, new Path(args.dst, dst));
// skip file if it exceed file limit or size limit
skipfile |=
fileCount == args.filelimit || byteCount + child.getLen() > args.sizelimit;
if (!skipfile) {
++fileCount;
byteCount += child.getLen();
if (LOG.isTraceEnabled()) {
LOG.trace("adding file " + child.getPath());
}
++cnsyncf;
cbsyncs += child.getLen();
if (cnsyncf > SYNC_FILE_MAX || cbsyncs > BYTES_PER_MAP) {
src_writer.sync();
dst_writer.sync();
cnsyncf = 0;
cbsyncs = 0L;
}
}
}
if (!skipfile) {
src_writer.append(
new LongWritable(child.isDir() ? 0 : child.getLen()), new FilePair(child, dst));
}
dst_writer.append(new Text(dst), new Text(child.getPath().toString()));
}
if (cur.isDir()) {
String dst = makeRelative(root, cur.getPath());
dir_writer.append(new Text(dst), new FilePair(cur, dst));
if (++dirsyn > SYNC_FILE_MAX) {
dirsyn = 0;
dir_writer.sync();
}
}
}
}
} finally {
checkAndClose(src_writer);
checkAndClose(dst_writer);
checkAndClose(dir_writer);
}
FileStatus dststatus = null;
try {
dststatus = dstfs.getFileStatus(args.dst);
} catch (FileNotFoundException fnfe) {
LOG.info(args.dst + " does not exist.");
}
// create dest path dir if copying > 1 file
if (dststatus == null) {
if (srcCount > 1 && !dstfs.mkdirs(args.dst)) {
throw new IOException("Failed to create" + args.dst);
}
}
final Path sorted = new Path(jobDirectory, "_distcp_sorted");
checkDuplication(jobfs, dstfilelist, sorted, conf);
if (dststatus != null && args.flags.contains(Options.DELETE)) {
deleteNonexisting(dstfs, dststatus, sorted, jobfs, jobDirectory, jobConf, conf);
}
Path tmpDir =
new Path(
(dstExists && !dstIsDir) || (!dstExists && srcCount == 1)
? args.dst.getParent()
: args.dst,
"_distcp_tmp_" + randomId);
jobConf.set(TMP_DIR_LABEL, tmpDir.toUri().toString());
LOG.info("srcCount=" + srcCount);
jobConf.setInt(SRC_COUNT_LABEL, srcCount);
jobConf.setLong(TOTAL_SIZE_LABEL, byteCount);
setMapCount(byteCount, jobConf);
}
/**
* Method to go though the HDFS filesystem in a DFS to find all files
*
* <p>fs:FileSystem object from HDFS minDate: Oldest date for files to be backed up maxDate:Newest
* date for files to be backed up p:Path in HDFS to look for files pathList:Will be filled with
* all files in p hmTimestamps: hashmap of timestamps for later sorting
*/
public void checkDir(
FileSystem fs,
long minDate,
long maxDate,
Path p,
ArrayList<Path> pathList,
HashMap<Path, Long> hmTimestamps) {
long tmpDate;
FileStatus[] fStat;
try {
String sPath = p.toUri().getPath();
// If this is a directory
if (fs.getFileStatus(p).isDir()) {
// ignore certain directories
if ("dfstmp".equals(p.getName())
|| "tmp".equals(p.getName())
|| "jobtracker".equals(p.getName())
|| sPath.startsWith("/mapred")
|| "ops".equals(p.getName())
|| p.getName().startsWith("_distcp_logs")) {
return;
}
// dump the mkdir and chmod commands for this
// directory -- skip root directory only
{
FileStatus stat = fs.getFileStatus(p);
if (!sPath.equals("/")) {
m_wrMkdirs.println("hadoop fs -mkdir " + sPath);
}
m_wrChmods.println(
"hadoop fs -chown " + stat.getOwner() + ":" + stat.getGroup() + " " + sPath);
Short sh = new Short(stat.getPermission().toShort());
m_wrChmods.println(
"hadoop fs -chmod " + Long.toOctalString(sh.longValue()) + " " + sPath);
}
fStat = fs.listStatus(p);
// Do a recursive call to all elements
for (int i = 0; i < fStat.length; i++) {
checkDir(fs, minDate, maxDate, fStat[i].getPath(), pathList, hmTimestamps);
}
} else {
// If not a directory then we've found a file
// ignore crc files
if (p.getName().endsWith(".crc")) {
return;
}
// ignore other files
if (sPath.startsWith("/user/oozie/etl/workflows/")) {
return;
}
// try to get the table name from the path. There are
// various types of tables, from those replicated from
// another database to regular hive tables to
// partitioned hive tables. We use table names to
// both exclude some from the backup, and for the rest
// to dump out the schema and partition name.
if (m_ignoreTables != null && m_ignoreTables.doIgnoreFile(sPath)) {
m_nIgnoredTables++;
if (m_nIgnoredTables < 5) {
System.out.println("Skipping ignore-table file: " + sPath);
} else if (m_nIgnoredTables == 5) {
System.out.println("(...not showing other skipped tables...)");
}
return;
}
FileStatus stat = fs.getFileStatus(p);
tmpDate = stat.getModificationTime() / 1000;
// store the chmods/chowns for all files
m_wrChmods.println(
"hadoop fs -chown " + stat.getOwner() + ":" + stat.getGroup() + " " + sPath);
m_wrChmods.println("hadoop fs -chmod " + stat.getPermission().toShort() + " " + sPath);
// check dates. is it too young?
if (tmpDate < minDate) {
return;
}
// is the file too recent?
if (tmpDate > maxDate) {
// System.out.println("file too recent: " + sPath);
return;
}
// file timestamp is ok
pathList.add(p);
hmTimestamps.put(p, new Long(tmpDate));
// store info about total bytes neeed to backup
m_nTotalBytes += fs.getContentSummary(p).getLength();
}
} catch (IOException e) {
System.err.println("ERROR: could not open " + p + ": " + e);
// System.exit(1) ;
}
}
/**
* Method to move files from HDFS to local filesystem
*
* <p>localPath: Path on the machines filesystem fs:FileSystem object from HDFS pathList:List of
* paths for files that might need to be backed up size:max size in bytes to be backed up
*
* <p>ReturnsDate of the last files backed up if reached size limit, else, zero
*/
public long backupFiles(
String localPath, String preservePath, FileSystem fs, ArrayList<Path> pathList, long size) {
Path fsPath;
long tmpSize = 0;
long tmpDate = 0;
// Start iterating over all paths
for (Path hdfsPath : pathList) {
try {
long nFileSize = fs.getContentSummary(hdfsPath).getLength();
tmpSize = tmpSize + nFileSize;
if ((tmpSize <= size) || (size == 0)) {
FileStatus stat = fs.getFileStatus(hdfsPath);
System.err.println(
"File "
+ hdfsPath.toUri().getPath()
+ " "
+ nFileSize
+ " bytes, "
+ "perms: "
+ stat.getOwner()
+ "/"
+ stat.getGroup()
+ ", "
+ stat.getPermission().toString());
tmpDate = stat.getModificationTime() / 1000;
String sFsPath = localPath + hdfsPath.toUri().getPath();
fsPath = new Path(sFsPath);
File f = new File(sFsPath);
// COMMENTED OUT: until a few backup cycles run
// and the mtime gets in fact set on all copied
// files.
//
// ignore it if the file exists and has the same mtime
// if (f.exists() && f.isFile() && f.lastModified() == stat.getModificationTime())
// {
// System.out.println("no need to backup " + f.toString() + ", mtime matches hdfs");
// continue;
// }
if (false == m_bDryRun) {
// check if we need to back up the local file
// (not directory), if it already exists.
if (f.exists() && f.isFile()) {
// ignore files with substrings in the
// no-preserve file
if (true == doPreserveFile(sFsPath)) {
// move it to the backup path
String sNewPath = preservePath + hdfsPath.toUri().getPath();
File newFile = new File(sNewPath);
// create directory structure for new file?
if (false == newFile.getParentFile().exists()) {
if (false == newFile.getParentFile().mkdirs()) {
System.err.println("Failed to mkdirs " + newFile.getParentFile().toString());
System.exit(1);
}
}
// rename existing file to new location
if (false == f.renameTo(newFile)) {
System.err.println(
"Failed to renameTo " + f.toString() + " to " + newFile.toString());
System.exit(1);
}
System.out.println("preserved " + f.toString() + " into " + newFile.toString());
} else {
System.out.println("skipped preservation of " + f.toString());
}
}
// copy from hdfs to local filesystem
fs.copyToLocalFile(hdfsPath, fsPath);
// set the mtime to match hdfs file
f.setLastModified(stat.getModificationTime());
// compare checksums on both files
compareChecksums(fs, hdfsPath, sFsPath);
}
// don't print the progress after every file -- go
// by at least 1% increments
long nPercentDone = (long) (100 * tmpSize / m_nTotalBytes);
if (nPercentDone > m_nLastPercentBytesDone) {
System.out.println(
"progress: copied "
+ prettyPrintBytes(tmpSize)
+ ", "
+ nPercentDone
+ "% done"
+ ", tstamp="
+ tmpDate);
m_nLastPercentBytesDone = nPercentDone;
}
if (m_nSleepSeconds > 0) {
try {
Thread.sleep(1000 * m_nSleepSeconds);
} catch (Exception e2) {
// ignore
}
}
} else {
return tmpDate;
}
} catch (IOException e) {
System.err.println("FATAL ERROR: Something wrong with the file");
System.err.println(e);
System.out.println(tmpDate);
System.exit(1);
return 0;
}
}
return 0;
}
public void doTestTextBatchAppend() throws Exception {
LOG.debug("Starting...");
final long rollCount = 10;
final long batchSize = 2;
final String fileName = "PageView";
String newPath = testPath + "/singleTextBucket";
int totalEvents = 0;
int i = 1, j = 1;
// clear the test directory
Configuration conf = new Configuration();
FileSystem fs = FileSystem.get(conf);
Path dirPath = new Path(newPath);
fs.delete(dirPath, true);
fs.mkdirs(dirPath);
Context context = new Context();
context.put("hdfs.path", newPath);
context.put("hdfs.rollCount", String.valueOf(rollCount));
context.put("hdfs.batchSize", String.valueOf(batchSize));
context.put("hdfs.filePrefix", "pageview");
Channel channel = new MemoryChannel();
Configurables.configure(channel, context);
sink.setChannel(channel);
sink.start();
Calendar eventDate = Calendar.getInstance();
Date currentDate = new Date();
Map<String, String> header = new HashMap<String, String>();
header.put("topic", "PageView");
List<String> bodies = Lists.newArrayList();
// 将测试的事件推入到通道中
for (i = 1; i <= (rollCount * 10) / batchSize; i++) {
Transaction txn = channel.getTransaction();
txn.begin();
for (j = 1; j <= batchSize; j++) {
header.put("timestamp", String.valueOf(currentDate.getTime()));
Event event = new SimpleEvent();
eventDate.clear();
eventDate.set(2014, i, i, i, 0);
String body = "Test." + i + "." + j;
event.setHeaders(header);
event.setBody(body.getBytes());
bodies.add(body);
channel.put(event);
totalEvents++;
}
txn.commit();
txn.close();
// execute sink to process the events
sink.process();
}
sink.stop();
FileStatus[] dirStat = fs.listStatus(dirPath);
Path fList[] = FileUtil.stat2Paths(dirStat);
long expectedFiles = totalEvents / rollCount;
if (totalEvents % rollCount > 0) {
expectedFiles++;
}
Assert.assertEquals(
"num files wrong, found: " + Lists.newArrayList(fList), expectedFiles, fList.length);
// 检查所有写入文件的内容
verifyOutputTextFiles(fs, conf, dirPath.toUri().getPath(), fileName, bodies);
}