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);
}
/** Convert a Json map to a HdfsFileStatus object. */
public static HdfsFileStatus toFileStatus(final Map<?, ?> json, boolean includesType) {
if (json == null) {
return null;
}
final Map<?, ?> m =
includesType ? (Map<?, ?>) json.get(FileStatus.class.getSimpleName()) : json;
final String localName = (String) m.get("pathSuffix");
final PathType type = PathType.valueOf((String) m.get("type"));
final byte[] symlink =
type != PathType.SYMLINK ? null : DFSUtil.string2Bytes((String) m.get("symlink"));
final long len = (Long) m.get("length");
final String owner = (String) m.get("owner");
final String group = (String) m.get("group");
final FsPermission permission =
toFsPermission(
(String) m.get("permission"), (Boolean) m.get("aclBit"), (Boolean) m.get("encBit"));
final long aTime = (Long) m.get("accessTime");
final long mTime = (Long) m.get("modificationTime");
final long blockSize = (Long) m.get("blockSize");
final short replication = (short) (long) (Long) m.get("replication");
final long fileId =
m.containsKey("fileId") ? (Long) m.get("fileId") : INodeId.GRANDFATHER_INODE_ID;
Long childrenNumLong = (Long) m.get("childrenNum");
final int childrenNum = (childrenNumLong == null) ? -1 : childrenNumLong.intValue();
final byte storagePolicy =
m.containsKey("storagePolicy")
? (byte) (long) (Long) m.get("storagePolicy")
: BlockStoragePolicySuite.ID_UNSPECIFIED;
return new HdfsFileStatus(
len,
type == PathType.DIRECTORY,
replication,
blockSize,
mTime,
aTime,
permission,
owner,
group,
symlink,
DFSUtil.string2Bytes(localName),
fileId,
childrenNum,
null,
storagePolicy);
}
/**
* Parse a list of strings into longs.
*
* @param strs the list of strings to parse
* @return a list of longs that were parsed. same length as strs.
*/
private static long[] parseTimestamps(String[] strs) {
if (strs == null) {
return null;
}
long[] result = new long[strs.length];
for (int i = 0; i < strs.length; ++i) {
result[i] = Long.parseLong(strs[i]);
}
return result;
}
/** Sets up configuration based on params */
private static boolean setup(Hashtable<String, String> curConf, Configuration argConf) {
if (argConf.get("file") == null) {
logger.fatal("Missing file parameter");
System.exit(1);
}
if (argConf.get("hdfs_base_path") == null) {
logger.fatal("Missing HDFS base path, check gestore-conf.xml");
System.exit(1);
}
if (argConf.get("hdfs_temp_path") == null) {
logger.fatal("Missing HDFS temp path, check gestore-conf.xml");
System.exit(1);
}
if (argConf.get("local_temp_path") == null) {
logger.fatal("Missing local temp path, check gestore-conf.xml");
System.exit(1);
}
// Input paramaters
curConf.put("run_id", argConf.get("run", ""));
curConf.put("task_id", argConf.get("task", ""));
curConf.put("file_id", argConf.get("file"));
curConf.put("local_path", argConf.get("path", ""));
curConf.put("type", argConf.get("type", "l2r"));
curConf.put("timestamp_start", argConf.get("timestamp_start", "1"));
curConf.put(
"timestamp_stop", argConf.get("timestamp_stop", Integer.toString(Integer.MAX_VALUE)));
curConf.put("delimiter", argConf.get("regex", "ID=.*"));
curConf.put("taxon", argConf.get("taxon", "all"));
curConf.put("intermediate", argConf.get("full_run", "false"));
curConf.put("quick_add", argConf.get("quick_add", "false"));
Boolean full_run = curConf.get("intermediate").matches("(?i).*true.*");
curConf.put("format", argConf.get("format", "unknown"));
curConf.put("split", argConf.get("split", "1"));
curConf.put("copy", argConf.get("copy", "true"));
// Constants
curConf.put("base_path", argConf.get("hdfs_base_path"));
curConf.put("temp_path", argConf.get("hdfs_temp_path"));
curConf.put("local_temp_path", argConf.get("local_temp_path"));
curConf.put("db_name_files", argConf.get("hbase_file_table"));
curConf.put("db_name_runs", argConf.get("hbase_run_table"));
curConf.put("db_name_updates", argConf.get("hbase_db_update_table"));
// Timestamps
Date currentTime = new Date();
Date endDate = new Date(new Long(curConf.get("timestamp_stop")));
curConf.put("timestamp_real", Long.toString(currentTime.getTime()));
return true;
}
/** Find the corresponding meta data file from a given block file */
private static long parseGenerationStamp(File blockFile, File metaFile) throws IOException {
String metaname = metaFile.getName();
String gs =
metaname.substring(
blockFile.getName().length() + 1, metaname.length() - METADATA_EXTENSION.length());
try {
return Long.parseLong(gs);
} catch (NumberFormatException nfe) {
throw (IOException)
new IOException("blockFile=" + blockFile + ", metaFile=" + metaFile).initCause(nfe);
}
}
/**
* Find the metadata file for the specified block file. Return the generation stamp from the
* name of the metafile.
*/
long getGenerationStampFromFile(File[] listdir, File blockFile) {
String blockName = blockFile.getName();
for (int j = 0; j < listdir.length; j++) {
String path = listdir[j].getName();
if (!path.startsWith(blockName)) {
continue;
}
String[] vals = path.split("_");
if (vals.length != 3) { // blk, blkid, genstamp.meta
continue;
}
String[] str = vals[2].split("\\.");
if (str.length != 2) {
continue;
}
return Long.parseLong(str[0]);
}
DataNode.LOG.warn("Block " + blockFile + " does not have a metafile!");
return Block.GRANDFATHER_GENERATION_STAMP;
}
/**
* Update the maps baseDirSize and baseDirNumberSubDir when adding cache.
*
* @param cacheStatus cache status of the cache is added
*/
private static void addCacheInfoUpdate(CacheStatus cacheStatus) {
long cacheSize = cacheStatus.size;
synchronized (baseDirSize) {
Long dirSize = baseDirSize.get(cacheStatus.getBaseDir());
if (dirSize == null) {
dirSize = Long.valueOf(cacheSize);
} else {
dirSize += cacheSize;
}
baseDirSize.put(cacheStatus.getBaseDir(), dirSize);
}
synchronized (baseDirNumberSubDir) {
Integer dirSubDir = baseDirNumberSubDir.get(cacheStatus.getBaseDir());
if (dirSubDir == null) {
dirSubDir = 1;
} else {
dirSubDir += 1;
}
baseDirNumberSubDir.put(cacheStatus.getBaseDir(), dirSubDir);
}
}
/**
* Get the locally cached file or archive; it could either be previously cached (and valid) or
* copy it from the {@link FileSystem} now.
*
* @param cache the cache to be localized, this should be specified as new
* URI(hdfs://hostname:port/absolute_path_to_file#LINKNAME). If no schema or hostname:port is
* provided the file is assumed to be in the filesystem being used in the Configuration
* @param conf The Confguration file which contains the filesystem
* @param subDir The sub cache Dir where you want to localize the files/archives
* @param fileStatus The file status on the dfs.
* @param isArchive if the cache is an archive or a file. In case it is an archive with a .zip or
* .jar or .tar or .tgz or .tar.gz extension it will be unzipped/unjarred/untarred
* automatically and the directory where the archive is unzipped/unjarred/untarred is returned
* as the Path. In case of a file, the path to the file is returned
* @param confFileStamp this is the hdfs file modification timestamp to verify that the file to be
* cached hasn't changed since the job started
* @param fileLength this is the length of the cache file
* @param currentWorkDir this is the directory where you would want to create symlinks for the
* locally cached files/archives
* @param honorSymLinkConf if this is false, then the symlinks are not created even if conf says
* so (this is required for an optimization in task launches
* @param lDirAllocator LocalDirAllocator of the tracker
* @return the path to directory where the archives are unjarred in case of archives, the path to
* the file where the file is copied locally
* @throws IOException
*/
private static Path getLocalCache(
URI cache,
Configuration conf,
Path subDir,
FileStatus fileStatus,
boolean isArchive,
long confFileStamp,
long fileLength,
Path currentWorkDir,
boolean honorSymLinkConf,
MRAsyncDiskService asyncDiskService,
LocalDirAllocator lDirAllocator)
throws IOException {
String key = getKey(cache, conf, confFileStamp);
CacheStatus lcacheStatus;
Path localizedPath;
synchronized (cachedArchives) {
lcacheStatus = cachedArchives.get(key);
if (lcacheStatus == null) {
// was never localized
Path uniqueParentDir = new Path(subDir, String.valueOf(random.nextLong()));
String cachePath = new Path(uniqueParentDir, makeRelative(cache, conf)).toString();
Path localPath = lDirAllocator.getLocalPathForWrite(cachePath, fileLength, conf);
lcacheStatus =
new CacheStatus(
new Path(localPath.toString().replace(cachePath, "")), localPath, uniqueParentDir);
cachedArchives.put(key, lcacheStatus);
}
lcacheStatus.refcount++;
}
boolean initSuccessful = false;
try {
synchronized (lcacheStatus) {
if (!lcacheStatus.isInited()) {
localizedPath = localizeCache(conf, cache, confFileStamp, lcacheStatus, isArchive);
lcacheStatus.initComplete();
} else {
if (fileStatus != null) {
localizedPath =
checkCacheStatusValidity(
conf, cache, confFileStamp, lcacheStatus, fileStatus, isArchive);
} else {
// if fileStatus is null, then the md5 must be correct
// so there is no need to check for cache validity
localizedPath = lcacheStatus.localizedLoadPath;
}
}
createSymlink(conf, cache, lcacheStatus, isArchive, currentWorkDir, honorSymLinkConf);
}
// try deleting stuff if you can
long size = 0;
int numberSubDir = 0;
synchronized (lcacheStatus) {
synchronized (baseDirSize) {
Long get = baseDirSize.get(lcacheStatus.getBaseDir());
if (get != null) {
size = get.longValue();
} else {
LOG.warn("Cannot find size of baseDir: " + lcacheStatus.getBaseDir());
}
}
synchronized (baseDirNumberSubDir) {
Integer get = baseDirNumberSubDir.get(lcacheStatus.getBaseDir());
if (get != null) {
numberSubDir = get.intValue();
} else {
LOG.warn("Cannot find subdirectories limit of baseDir: " + lcacheStatus.getBaseDir());
}
}
}
// setting the cache size to a default of 10GB
long allowedSize = conf.getLong("local.cache.size", DEFAULT_CACHE_SIZE);
long allowedNumberSubDir =
conf.getLong("local.cache.numbersubdir", DEFAULT_CACHE_SUBDIR_LIMIT);
if (allowedSize < size || allowedNumberSubDir < numberSubDir) {
// try some cache deletions
LOG.debug(
"Start deleting released cache because"
+ " [size, allowedSize, numberSubDir, allowedNumberSubDir] ="
+ " ["
+ size
+ ", "
+ allowedSize
+ ", "
+ numberSubDir
+ ", "
+ allowedNumberSubDir
+ "]");
deleteCache(conf, asyncDiskService);
}
initSuccessful = true;
return localizedPath;
} finally {
if (!initSuccessful) {
synchronized (cachedArchives) {
lcacheStatus.refcount--;
}
}
}
}
/**
* 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) ;
}
}
public int run(String[] args) throws Exception {
// printUsage();
/*
* SETUP
*/
Configuration argConf = getConf();
Hashtable<String, String> confArg = new Hashtable<String, String>();
setup(confArg, argConf);
Date currentTime = new Date();
Date endDate = new Date(new Long(confArg.get("timestamp_stop")));
Boolean full_run = confArg.get("intermediate").matches("(?i).*true.*");
Boolean quick_add = confArg.get("quick_add").matches("(?i).*true.*");
logger.info("Running GeStore");
// ZooKeeper setup
Configuration config = HBaseConfiguration.create();
zkWatcher = new ZooKeeperWatcher(config, "Testing", new HBaseAdmin(config));
zkInstance =
new ZooKeeper(
ZKConfig.getZKQuorumServersString(config),
config.getInt("zookeeper.session.timeout", -1),
zkWatcher);
if (!confArg.get("task_id").isEmpty()) {
confArg.put("temp_path", confArg.get("temp_path") + confArg.get("task_id"));
}
String lockRequest = confArg.get("file_id");
if (!confArg.get("run_id").isEmpty())
lockRequest = lockRequest + "_" + confArg.get("run_id") + "_";
if (!confArg.get("task_id").isEmpty())
lockRequest = lockRequest + "_" + confArg.get("task_id") + "_";
// Get type of movement
toFrom type_move = checkArgs(confArg);
if (type_move == toFrom.LOCAL2REMOTE && !confArg.get("format").equals("unknown")) {
List<String> arguments = new ArrayList<String>();
arguments.add("-Dinput=" + confArg.get("local_path"));
arguments.add("-Dtable=" + confArg.get("file_id"));
arguments.add("-Dtimestamp=" + confArg.get("timestamp_stop"));
arguments.add("-Dtype=" + confArg.get("format"));
arguments.add("-Dtarget_dir=" + confArg.get("base_path") + "_" + confArg.get("file_id"));
arguments.add("-Dtemp_hdfs_path=" + confArg.get("temp_path"));
arguments.add("-Drun_id=" + confArg.get("run_id"));
if (!confArg.get("run_id").isEmpty()) arguments.add("-Drun_id=" + confArg.get("run_id"));
if (!confArg.get("task_id").isEmpty()) arguments.add("-Dtask_id=" + confArg.get("task_id"));
if (quick_add) arguments.add("-Dquick_add=" + confArg.get("quick_add"));
String lockName = lock(lockRequest);
String[] argumentString = arguments.toArray(new String[arguments.size()]);
adddb.main(argumentString);
unlock(lockName);
System.exit(0);
}
// Database registration
dbutil db_util = new dbutil(config);
db_util.register_database(confArg.get("db_name_files"), true);
db_util.register_database(confArg.get("db_name_runs"), true);
db_util.register_database(confArg.get("db_name_updates"), true);
FileSystem hdfs = FileSystem.get(config);
FileSystem localFS = FileSystem.getLocal(config);
// Get source type
confArg.put("source", getSource(db_util, confArg.get("db_name_files"), confArg.get("file_id")));
confArg.put(
"database", isDatabase(db_util, confArg.get("db_name_files"), confArg.get("file_id")));
if (!confArg.get("source").equals("local")
&& type_move == toFrom.REMOTE2LOCAL
&& !confArg.get("timestamp_stop").equals(Integer.toString(Integer.MAX_VALUE))) {
confArg.put("timestamp_stop", Long.toString(latestVersion(confArg, db_util)));
}
/*
* Get previous timestamp
*/
Get run_id_get = new Get(confArg.get("run_id").getBytes());
Result run_get = db_util.doGet(confArg.get("db_name_runs"), run_id_get);
KeyValue run_file_prev =
run_get.getColumnLatest(
"d".getBytes(), (confArg.get("file_id") + "_db_timestamp").getBytes());
String last_timestamp = new String("0");
if (null != run_file_prev && !confArg.get("source").equals("local")) {
long last_timestamp_real = run_file_prev.getTimestamp();
Long current_timestamp = new Long(confArg.get("timestamp_real"));
if ((current_timestamp - last_timestamp_real) > 36000) {
last_timestamp = new String(run_file_prev.getValue());
Integer lastTimestamp = new Integer(last_timestamp);
lastTimestamp += 1;
last_timestamp = lastTimestamp.toString();
logger.info("Last timestamp: " + last_timestamp + " End data: " + endDate);
Date last_run = new Date(run_file_prev.getTimestamp());
if (last_run.before(endDate) && !full_run) {
confArg.put("timestamp_start", last_timestamp);
}
}
}
Integer tse = new Integer(confArg.get("timestamp_stop"));
Integer tss = new Integer(confArg.get("timestamp_start"));
if (tss > tse) {
logger.info("No new version of requested file.");
return 0;
}
/*
* Generate file
*/
String lockName = lock(lockRequest);
Get file_id_get = new Get(confArg.get("file_id").getBytes());
Result file_get = db_util.doGet(confArg.get("db_name_files"), file_id_get);
if (!file_get.isEmpty()) {
boolean found =
hasFile(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
getFullPath(confArg));
if (confArg.get("source").equals("fullfile")) {
found = false;
}
String filenames_put =
getFileNames(
db_util, confArg.get("db_name_files"), confArg.get("file_id"), getFullPath(confArg));
// Filename not found in file database
if (!found && type_move == toFrom.REMOTE2LOCAL) {
if (!confArg.get("source").equals("local")) {
// Generate intermediate file
if (getFile(hdfs, confArg, db_util) == null) {
unlock(lockName);
return 1;
}
// Put generated file into file database
if (!confArg.get("format").equals("fullfile")) {
putFileEntry(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
confArg.get("full_file_name"),
confArg.get("source"));
}
} else {
logger.warn("Remote file not found, and cannot be generated! File: " + confArg);
unlock(lockName);
return 1;
}
}
} else {
if (type_move == toFrom.REMOTE2LOCAL) {
logger.warn("Remote file not found, and cannot be generated.");
unlock(lockName);
return 1;
}
}
/*
* Copy file
* Update tables
*/
if (type_move == toFrom.LOCAL2REMOTE) {
if (!confArg.get("format").equals("fullfile")) {
putFileEntry(
db_util,
hdfs,
confArg.get("db_name_files"),
confArg.get("file_id"),
getFullPath(confArg),
confArg.get("source"));
}
putRunEntry(
db_util,
confArg.get("db_name_runs"),
confArg.get("run_id"),
confArg.get("file_id"),
confArg.get("type"),
confArg.get("timestamp_real"),
confArg.get("timestamp_stop"),
getFullPath(confArg),
confArg.get("delimiter"));
hdfs.copyFromLocalFile(new Path(confArg.get("local_path")), new Path(getFullPath(confArg)));
} else if (type_move == toFrom.REMOTE2LOCAL) {
FileStatus[] files = hdfs.globStatus(new Path(getFullPath(confArg) + "*"));
putRunEntry(
db_util,
confArg.get("db_name_runs"),
confArg.get("run_id"),
confArg.get("file_id"),
confArg.get("type"),
confArg.get("timestamp_real"),
confArg.get("timestamp_stop"),
getFullPath(confArg),
confArg.get("delimiter"));
unlock(lockName);
for (FileStatus file : files) {
Path cur_file = file.getPath();
Path cur_local_path =
new Path(new String(confArg.get("local_path") + confArg.get("file_id")));
String suffix = getSuffix(getFileName(confArg), cur_file.getName());
if (suffix.length() > 0) {
cur_local_path = cur_local_path.suffix(new String("." + suffix));
}
if (confArg.get("copy").equals("true")) {
String crc = hdfs.getFileChecksum(cur_file).toString();
if (checksumLocalTest(cur_local_path, crc)) {
continue;
} else {
hdfs.copyToLocalFile(cur_file, cur_local_path);
writeChecksum(cur_local_path, crc);
}
} else {
System.out.println(cur_local_path + "\t" + cur_file);
}
}
}
unlock(lockName);
return 0;
}
// Information needed to get a single file:
// BASE_PATH, FILE_ID, TIMESTAMP_START, TIMESTAMP_STOP, SOURCE, FILESYSTEM
private static Vector<Path> getFile(
FileSystem fs, Hashtable<String, String> config, dbutil db_util) throws Exception {
Long latestVersion = latestVersion(config, db_util);
try {
config.put("timestamp_start", config.get("timestamp_start"));
config.put("timestamp_real", latestVersion.toString());
config.put("timestamp_stop", latestVersion.toString());
} catch (Exception E) {
logger.error("Tryign to get file that is impossible to generate: " + getFullPath(config));
return null;
}
if (Integer.parseInt(config.get("timestamp_start"))
> Integer.parseInt(config.get("timestamp_stop"))) {
return null;
}
logger.debug(
"Getting DB for timestamp "
+ config.get("timestamp_start")
+ " to "
+ config.get("timestamp_stop"));
String final_result = getFullPath(config);
String temp_path_base =
config.get("local_temp_path")
+ "_"
+ config.get("task_id")
+ "_"
+ config.get("run_id")
+ "/";
Path newPath = new Path(final_result + "*");
Vector<Path> ret_path = new Vector<Path>();
String lockName = lock(final_result.replaceAll("/", "_"));
if (fs.globStatus(newPath).length != 0) {
ret_path.add(newPath);
unlock(lockName);
config.put("full_file_name", final_result);
return ret_path;
} else {
if (!config.get("source").equals("local")) {
config.put("temp_path_base", temp_path_base);
config.put("timestamp_start", config.get("timestamp_start"));
config.put("timestamp_real", latestVersion.toString());
config.put("timestamp_stop", latestVersion.toString());
Class<?> sourceClass =
Class.forName("org.gestore.plugin.source." + config.get("source") + "Source");
Method process_data = sourceClass.getMethod("process", Hashtable.class, FileSystem.class);
Object processor = sourceClass.newInstance();
Object retVal;
try {
retVal = process_data.invoke(processor, config, fs);
} catch (InvocationTargetException E) {
Throwable exception = E.getTargetException();
logger.error("Unable to call method in child class: " + exception.toString());
exception.printStackTrace(System.out);
unlock(lockName);
return null;
}
FileStatus[] files = (FileStatus[]) retVal;
if (files == null) {
logger.error("Error getting files, no files returned");
return null;
}
for (FileStatus file : files) {
Path cur_file = file.getPath();
Path cur_local_path = new Path(temp_path_base + config.get("file_id"));
String suffix = getSuffix(config.get("file_id"), cur_file.getName());
cur_local_path = cur_local_path.suffix(suffix);
Path res_path = new Path(new String(final_result + suffix));
logger.debug("Moving file" + cur_file.toString() + " to " + res_path.toString());
if (config.get("copy").equals("true")) {
fs.moveFromLocalFile(cur_file, res_path);
} else {
fs.rename(cur_file, res_path);
}
}
config.put("full_file_name", final_result);
}
}
unlock(lockName);
return ret_path;
}
@Override
public final void run() {
try {
// before preparing the job localize
// all the archives
TaskAttemptID taskid = t.getTaskID();
LocalDirAllocator lDirAlloc = new LocalDirAllocator("mapred.local.dir");
File jobCacheDir = null;
if (conf.getJar() != null) {
jobCacheDir = new File(new Path(conf.getJar()).getParent().toString());
}
File workDir =
new File(
lDirAlloc
.getLocalPathToRead(
TaskTracker.getJobCacheSubdir()
+ Path.SEPARATOR
+ t.getJobID()
+ Path.SEPARATOR
+ t.getTaskID()
+ Path.SEPARATOR
+ MRConstants.WORKDIR,
conf)
.toString());
URI[] archives = DistributedCache.getCacheArchives(conf);
URI[] files = DistributedCache.getCacheFiles(conf);
FileStatus fileStatus;
FileSystem fileSystem;
Path localPath;
String baseDir;
if ((archives != null) || (files != null)) {
if (archives != null) {
String[] archivesTimestamps = DistributedCache.getArchiveTimestamps(conf);
Path[] p = new Path[archives.length];
for (int i = 0; i < archives.length; i++) {
fileSystem = FileSystem.get(archives[i], conf);
fileStatus = fileSystem.getFileStatus(new Path(archives[i].getPath()));
String cacheId = DistributedCache.makeRelative(archives[i], conf);
String cachePath = TaskTracker.getCacheSubdir() + Path.SEPARATOR + cacheId;
if (lDirAlloc.ifExists(cachePath, conf)) {
localPath = lDirAlloc.getLocalPathToRead(cachePath, conf);
} else {
localPath = lDirAlloc.getLocalPathForWrite(cachePath, fileStatus.getLen(), conf);
}
baseDir = localPath.toString().replace(cacheId, "");
p[i] =
DistributedCache.getLocalCache(
archives[i],
conf,
new Path(baseDir),
fileStatus,
true,
Long.parseLong(archivesTimestamps[i]),
new Path(workDir.getAbsolutePath()),
false);
}
DistributedCache.setLocalArchives(conf, stringifyPathArray(p));
}
if ((files != null)) {
String[] fileTimestamps = DistributedCache.getFileTimestamps(conf);
Path[] p = new Path[files.length];
for (int i = 0; i < files.length; i++) {
fileSystem = FileSystem.get(files[i], conf);
fileStatus = fileSystem.getFileStatus(new Path(files[i].getPath()));
String cacheId = DistributedCache.makeRelative(files[i], conf);
String cachePath = TaskTracker.getCacheSubdir() + Path.SEPARATOR + cacheId;
if (lDirAlloc.ifExists(cachePath, conf)) {
localPath = lDirAlloc.getLocalPathToRead(cachePath, conf);
} else {
localPath = lDirAlloc.getLocalPathForWrite(cachePath, fileStatus.getLen(), conf);
}
baseDir = localPath.toString().replace(cacheId, "");
p[i] =
DistributedCache.getLocalCache(
files[i],
conf,
new Path(baseDir),
fileStatus,
false,
Long.parseLong(fileTimestamps[i]),
new Path(workDir.getAbsolutePath()),
false);
}
DistributedCache.setLocalFiles(conf, stringifyPathArray(p));
}
Path localTaskFile = new Path(t.getJobFile());
FileSystem localFs = FileSystem.getLocal(conf);
localFs.delete(localTaskFile, true);
OutputStream out = localFs.create(localTaskFile);
try {
conf.writeXml(out);
} finally {
out.close();
}
}
if (!prepare()) {
return;
}
String sep = System.getProperty("path.separator");
StringBuffer classPath = new StringBuffer();
// start with same classpath as parent process
classPath.append(System.getProperty("java.class.path"));
classPath.append(sep);
if (!workDir.mkdirs()) {
if (!workDir.isDirectory()) {
LOG.fatal("Mkdirs failed to create " + workDir.toString());
}
}
String jar = conf.getJar();
if (jar != null) {
// if jar exists, it into workDir
File[] libs = new File(jobCacheDir, "lib").listFiles();
if (libs != null) {
for (int i = 0; i < libs.length; i++) {
classPath.append(sep); // add libs from jar to classpath
classPath.append(libs[i]);
}
}
classPath.append(sep);
classPath.append(new File(jobCacheDir, "classes"));
classPath.append(sep);
classPath.append(jobCacheDir);
}
// include the user specified classpath
// archive paths
Path[] archiveClasspaths = DistributedCache.getArchiveClassPaths(conf);
if (archiveClasspaths != null && archives != null) {
Path[] localArchives = DistributedCache.getLocalCacheArchives(conf);
if (localArchives != null) {
for (int i = 0; i < archives.length; i++) {
for (int j = 0; j < archiveClasspaths.length; j++) {
if (archives[i].getPath().equals(archiveClasspaths[j].toString())) {
classPath.append(sep);
classPath.append(localArchives[i].toString());
}
}
}
}
}
// file paths
Path[] fileClasspaths = DistributedCache.getFileClassPaths(conf);
if (fileClasspaths != null && files != null) {
Path[] localFiles = DistributedCache.getLocalCacheFiles(conf);
if (localFiles != null) {
for (int i = 0; i < files.length; i++) {
for (int j = 0; j < fileClasspaths.length; j++) {
if (files[i].getPath().equals(fileClasspaths[j].toString())) {
classPath.append(sep);
classPath.append(localFiles[i].toString());
}
}
}
}
}
classPath.append(sep);
classPath.append(workDir);
// Build exec child jmv args.
Vector<String> vargs = new Vector<String>(8);
File jvm = // use same jvm as parent
new File(new File(System.getProperty("java.home"), "bin"), "java");
vargs.add(jvm.toString());
// Add child (task) java-vm options.
//
// The following symbols if present in mapred.child.java.opts value are
// replaced:
// + @taskid@ is interpolated with value of TaskID.
// Other occurrences of @ will not be altered.
//
// Example with multiple arguments and substitutions, showing
// jvm GC logging, and start of a passwordless JVM JMX agent so can
// connect with jconsole and the likes to watch child memory, threads
// and get thread dumps.
//
// <property>
// <name>mapred.child.java.opts</name>
// <value>-verbose:gc -Xloggc:/tmp/@[email protected] \
// -Dcom.sun.management.jmxremote.authenticate=false \
// -Dcom.sun.management.jmxremote.ssl=false \
// </value>
// </property>
//
String javaOpts = conf.get("mapred.child.java.opts", "-Xmx200m");
javaOpts = javaOpts.replace("@taskid@", taskid.toString());
String[] javaOptsSplit = javaOpts.split(" ");
// Add java.library.path; necessary for loading native libraries.
//
// 1. To support native-hadoop library i.e. libhadoop.so, we add the
// parent processes' java.library.path to the child.
// 2. We also add the 'cwd' of the task to it's java.library.path to help
// users distribute native libraries via the DistributedCache.
// 3. The user can also specify extra paths to be added to the
// java.library.path via mapred.child.java.opts.
//
String libraryPath = System.getProperty("java.library.path");
if (libraryPath == null) {
libraryPath = workDir.getAbsolutePath();
} else {
libraryPath += sep + workDir;
}
boolean hasUserLDPath = false;
for (int i = 0; i < javaOptsSplit.length; i++) {
if (javaOptsSplit[i].startsWith("-Djava.library.path=")) {
javaOptsSplit[i] += sep + libraryPath;
hasUserLDPath = true;
break;
}
}
if (!hasUserLDPath) {
vargs.add("-Djava.library.path=" + libraryPath);
}
for (int i = 0; i < javaOptsSplit.length; i++) {
vargs.add(javaOptsSplit[i]);
}
// add java.io.tmpdir given by mapred.child.tmp
String tmp = conf.get("mapred.child.tmp", "./tmp");
Path tmpDir = new Path(tmp);
// if temp directory path is not absolute
// prepend it with workDir.
if (!tmpDir.isAbsolute()) {
tmpDir = new Path(workDir.toString(), tmp);
}
FileSystem localFs = FileSystem.getLocal(conf);
if (!localFs.mkdirs(tmpDir) && !localFs.getFileStatus(tmpDir).isDir()) {
throw new IOException("Mkdirs failed to create " + tmpDir.toString());
}
vargs.add("-Djava.io.tmpdir=" + tmpDir.toString());
// Add classpath.
vargs.add("-classpath");
vargs.add(classPath.toString());
// Setup the log4j prop
long logSize = TaskLog.getTaskLogLength(conf);
vargs.add(
"-Dhadoop.log.dir=" + new File(System.getProperty("hadoop.log.dir")).getAbsolutePath());
vargs.add("-Dhadoop.root.logger=INFO,TLA");
vargs.add("-Dhadoop.tasklog.taskid=" + taskid);
vargs.add("-Dhadoop.tasklog.totalLogFileSize=" + logSize);
if (conf.getProfileEnabled()) {
if (conf.getProfileTaskRange(t.isMapTask()).isIncluded(t.getPartition())) {
File prof = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.PROFILE);
vargs.add(String.format(conf.getProfileParams(), prof.toString()));
}
}
// Add main class and its arguments
vargs.add(Child.class.getName()); // main of Child
// pass umbilical address
InetSocketAddress address = tracker.getTaskTrackerReportAddress();
vargs.add(address.getAddress().getHostAddress());
vargs.add(Integer.toString(address.getPort()));
vargs.add(taskid.toString()); // pass task identifier
String pidFile = null;
if (tracker.isTaskMemoryManagerEnabled()) {
pidFile =
lDirAlloc
.getLocalPathForWrite(
(TaskTracker.getPidFilesSubdir() + Path.SEPARATOR + taskid), this.conf)
.toString();
}
// set memory limit using ulimit if feasible and necessary ...
String[] ulimitCmd = Shell.getUlimitMemoryCommand(conf);
List<String> setup = null;
if (ulimitCmd != null) {
setup = new ArrayList<String>();
for (String arg : ulimitCmd) {
setup.add(arg);
}
}
// Set up the redirection of the task's stdout and stderr streams
File stdout = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDOUT);
File stderr = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDERR);
stdout.getParentFile().mkdirs();
tracker.getTaskTrackerInstrumentation().reportTaskLaunch(taskid, stdout, stderr);
Map<String, String> env = new HashMap<String, String>();
StringBuffer ldLibraryPath = new StringBuffer();
ldLibraryPath.append(workDir.toString());
String oldLdLibraryPath = null;
oldLdLibraryPath = System.getenv("LD_LIBRARY_PATH");
if (oldLdLibraryPath != null) {
ldLibraryPath.append(sep);
ldLibraryPath.append(oldLdLibraryPath);
}
env.put("LD_LIBRARY_PATH", ldLibraryPath.toString());
jvmManager.launchJvm(
this,
jvmManager.constructJvmEnv(
setup, vargs, stdout, stderr, logSize, workDir, env, pidFile, conf));
synchronized (lock) {
while (!done) {
lock.wait();
}
}
tracker.getTaskTrackerInstrumentation().reportTaskEnd(t.getTaskID());
if (exitCodeSet) {
if (!killed && exitCode != 0) {
if (exitCode == 65) {
tracker.getTaskTrackerInstrumentation().taskFailedPing(t.getTaskID());
}
throw new IOException("Task process exit with nonzero status of " + exitCode + ".");
}
}
} catch (FSError e) {
LOG.fatal("FSError", e);
try {
tracker.fsError(t.getTaskID(), e.getMessage());
} catch (IOException ie) {
LOG.fatal(t.getTaskID() + " reporting FSError", ie);
}
} catch (Throwable throwable) {
LOG.warn(t.getTaskID() + " Child Error", throwable);
ByteArrayOutputStream baos = new ByteArrayOutputStream();
throwable.printStackTrace(new PrintStream(baos));
try {
tracker.reportDiagnosticInfo(t.getTaskID(), baos.toString());
} catch (IOException e) {
LOG.warn(t.getTaskID() + " Reporting Diagnostics", e);
}
} finally {
try {
URI[] archives = DistributedCache.getCacheArchives(conf);
URI[] files = DistributedCache.getCacheFiles(conf);
if (archives != null) {
for (int i = 0; i < archives.length; i++) {
DistributedCache.releaseCache(archives[i], conf);
}
}
if (files != null) {
for (int i = 0; i < files.length; i++) {
DistributedCache.releaseCache(files[i], conf);
}
}
} catch (IOException ie) {
LOG.warn("Error releasing caches : Cache files might not have been cleaned up");
}
tracker.reportTaskFinished(t.getTaskID(), false);
if (t.isMapTask()) {
tracker.addFreeMapSlot();
} else {
tracker.addFreeReduceSlot();
}
}
}
public String toString() {
return Long.toString(value);
}