/**
* Returns TRUE if the history dirs were created, FALSE if they could not be created because the
* FileSystem is not reachable or in safe mode and throws and exception otherwise.
*/
@VisibleForTesting
boolean tryCreatingHistoryDirs(boolean logWait) throws IOException {
boolean succeeded = true;
String doneDirPrefix = JobHistoryUtils.getConfiguredHistoryServerDoneDirPrefix(conf);
try {
doneDirPrefixPath = FileContext.getFileContext(conf).makeQualified(new Path(doneDirPrefix));
doneDirFc = FileContext.getFileContext(doneDirPrefixPath.toUri(), conf);
doneDirFc.setUMask(JobHistoryUtils.HISTORY_DONE_DIR_UMASK);
mkdir(
doneDirFc,
doneDirPrefixPath,
new FsPermission(JobHistoryUtils.HISTORY_DONE_DIR_PERMISSION));
} catch (ConnectException ex) {
if (logWait) {
/* LOG.info("Waiting for FileSystem at "+doneDirPrefixPath.toUri().getAuthority()+"to be available") */
LOG.waiting_for_filesystem_available(
String.valueOf(doneDirPrefixPath.toUri().getAuthority()))
.tag("methodCall")
.info();
}
succeeded = false;
} catch (IOException e) {
if (isBecauseSafeMode(e)) {
succeeded = false;
if (logWait) {
/* LOG.info("Waiting for FileSystem at "+doneDirPrefixPath.toUri().getAuthority()+"to be out of safe mode") */
LOG.waiting_for_filesystem_out_safe_mode(
String.valueOf(doneDirPrefixPath.toUri().getAuthority()))
.tag("methodCall")
.info();
}
} else {
throw new YarnRuntimeException(
"Error creating done directory: [" + doneDirPrefixPath + "]", e);
}
}
if (succeeded) {
String intermediateDoneDirPrefix =
JobHistoryUtils.getConfiguredHistoryIntermediateDoneDirPrefix(conf);
try {
intermediateDoneDirPath =
FileContext.getFileContext(conf).makeQualified(new Path(intermediateDoneDirPrefix));
intermediateDoneDirFc = FileContext.getFileContext(intermediateDoneDirPath.toUri(), conf);
mkdir(
intermediateDoneDirFc,
intermediateDoneDirPath,
new FsPermission(JobHistoryUtils.HISTORY_INTERMEDIATE_DONE_DIR_PERMISSIONS.toShort()));
} catch (ConnectException ex) {
succeeded = false;
if (logWait) {
/* LOG.info("Waiting for FileSystem at "+intermediateDoneDirPath.toUri().getAuthority()+"to be available") */
LOG.waiting_for_filesystem_available(
String.valueOf(intermediateDoneDirPath.toUri().getAuthority()))
.tag("methodCall")
.info();
}
} catch (IOException e) {
if (isBecauseSafeMode(e)) {
succeeded = false;
if (logWait) {
/* LOG.info("Waiting for FileSystem at "+intermediateDoneDirPath.toUri().getAuthority()+"to be out of safe mode") */
LOG.waiting_for_filesystem_out_safe_mode(
String.valueOf(intermediateDoneDirPath.toUri().getAuthority()))
.tag("methodCall")
.info();
}
} else {
throw new YarnRuntimeException(
"Error creating intermediate done directory: [" + intermediateDoneDirPath + "]", e);
}
}
}
return succeeded;
}
@Override
protected void serviceInit(Configuration conf) throws Exception {
this.conf = conf;
int serialNumberLowDigits = 3;
serialNumberFormat =
("%0" + (JobHistoryUtils.SERIAL_NUMBER_DIRECTORY_DIGITS + serialNumberLowDigits) + "d");
String doneDirPrefix = null;
doneDirPrefix = JobHistoryUtils.getConfiguredHistoryServerDoneDirPrefix(conf);
try {
doneDirPrefixPath = FileContext.getFileContext(conf).makeQualified(new Path(doneDirPrefix));
doneDirFc = FileContext.getFileContext(doneDirPrefixPath.toUri(), conf);
doneDirFc.setUMask(JobHistoryUtils.HISTORY_DONE_DIR_UMASK);
mkdir(
doneDirFc,
doneDirPrefixPath,
new FsPermission(JobHistoryUtils.HISTORY_DONE_DIR_PERMISSION));
} catch (IOException e) {
throw new YarnRuntimeException(
"Error creating done directory: [" + doneDirPrefixPath + "]", e);
}
String intermediateDoneDirPrefix = null;
intermediateDoneDirPrefix = JobHistoryUtils.getConfiguredHistoryIntermediateDoneDirPrefix(conf);
try {
intermediateDoneDirPath =
FileContext.getFileContext(conf).makeQualified(new Path(intermediateDoneDirPrefix));
intermediateDoneDirFc = FileContext.getFileContext(intermediateDoneDirPath.toUri(), conf);
mkdir(
intermediateDoneDirFc,
intermediateDoneDirPath,
new FsPermission(JobHistoryUtils.HISTORY_INTERMEDIATE_DONE_DIR_PERMISSIONS.toShort()));
} catch (IOException e) {
LOG.info("error creating done directory on dfs " + e);
throw new YarnRuntimeException(
"Error creating intermediate done directory: [" + intermediateDoneDirPath + "]", e);
}
this.aclsMgr = new JobACLsManager(conf);
maxHistoryAge =
conf.getLong(JHAdminConfig.MR_HISTORY_MAX_AGE_MS, JHAdminConfig.DEFAULT_MR_HISTORY_MAX_AGE);
jobListCache =
new JobListCache(
conf.getInt(
JHAdminConfig.MR_HISTORY_JOBLIST_CACHE_SIZE,
JHAdminConfig.DEFAULT_MR_HISTORY_JOBLIST_CACHE_SIZE),
maxHistoryAge);
serialNumberIndex =
new SerialNumberIndex(
conf.getInt(
JHAdminConfig.MR_HISTORY_DATESTRING_CACHE_SIZE,
JHAdminConfig.DEFAULT_MR_HISTORY_DATESTRING_CACHE_SIZE));
int numMoveThreads =
conf.getInt(
JHAdminConfig.MR_HISTORY_MOVE_THREAD_COUNT,
JHAdminConfig.DEFAULT_MR_HISTORY_MOVE_THREAD_COUNT);
ThreadFactory tf =
new ThreadFactoryBuilder().setNameFormat("MoveIntermediateToDone Thread #%d").build();
moveToDoneExecutor =
new ThreadPoolExecutor(
numMoveThreads,
numMoveThreads,
1,
TimeUnit.HOURS,
new LinkedBlockingQueue<Runnable>(),
tf);
super.serviceInit(conf);
}
/* (non-Javadoc)
* @see org.apache.hadoop.yarn.service.AbstractService#init(org.
* apache.hadoop.conf.Configuration)
* Initializes the FileSystem and Path objects for the log and done directories.
* Creates these directories if they do not already exist.
*/
@Override
public void init(Configuration conf) {
String stagingDirStr = null;
String doneDirStr = null;
String userDoneDirStr = null;
try {
stagingDirStr = JobHistoryUtils.getConfiguredHistoryStagingDirPrefix(conf);
doneDirStr = JobHistoryUtils.getConfiguredHistoryIntermediateDoneDirPrefix(conf);
userDoneDirStr = JobHistoryUtils.getHistoryIntermediateDoneDirForUser(conf);
} catch (IOException e) {
LOG.error("Failed while getting the configured log directories", e);
throw new YarnException(e);
}
// Check for the existence of the history staging dir. Maybe create it.
try {
stagingDirPath = FileSystem.get(conf).makeQualified(new Path(stagingDirStr));
stagingDirFS = FileSystem.get(stagingDirPath.toUri(), conf);
mkdir(
stagingDirFS,
stagingDirPath,
new FsPermission(JobHistoryUtils.HISTORY_STAGING_DIR_PERMISSIONS));
} catch (IOException e) {
LOG.error(
"Failed while checking for/creating history staging path: [" + stagingDirPath + "]", e);
throw new YarnException(e);
}
// Check for the existence of intermediate done dir.
Path doneDirPath = null;
try {
doneDirPath = FileSystem.get(conf).makeQualified(new Path(doneDirStr));
doneDirFS = FileSystem.get(doneDirPath.toUri(), conf);
// This directory will be in a common location, or this may be a cluster
// meant for a single user. Creating based on the conf. Should ideally be
// created by the JobHistoryServer or as part of deployment.
if (!doneDirFS.exists(doneDirPath)) {
if (JobHistoryUtils.shouldCreateNonUserDirectory(conf)) {
LOG.info(
"Creating intermediate history logDir: ["
+ doneDirPath
+ "] + based on conf. Should ideally be created by the JobHistoryServer: "
+ MRJobConfig.MR_AM_CREATE_JH_INTERMEDIATE_BASE_DIR);
mkdir(
doneDirFS,
doneDirPath,
new FsPermission(
JobHistoryUtils.HISTORY_INTERMEDIATE_DONE_DIR_PERMISSIONS.toShort()));
// TODO Temporary toShort till new FsPermission(FsPermissions)
// respects
// sticky
} else {
String message =
"Not creating intermediate history logDir: ["
+ doneDirPath
+ "] based on conf: "
+ MRJobConfig.MR_AM_CREATE_JH_INTERMEDIATE_BASE_DIR
+ ". Either set to true or pre-create this directory with"
+ " appropriate permissions";
LOG.error(message);
throw new YarnException(message);
}
}
} catch (IOException e) {
LOG.error(
"Failed checking for the existance of history intermediate "
+ "done directory: ["
+ doneDirPath
+ "]");
throw new YarnException(e);
}
// Check/create user directory under intermediate done dir.
try {
doneDirPrefixPath = FileSystem.get(conf).makeQualified(new Path(userDoneDirStr));
mkdir(
doneDirFS,
doneDirPrefixPath,
new FsPermission(JobHistoryUtils.HISTORY_INTERMEDIATE_USER_DIR_PERMISSIONS));
} catch (IOException e) {
LOG.error(
"Error creating user intermediate history done directory: [ " + doneDirPrefixPath + "]",
e);
throw new YarnException(e);
}
// Maximum number of unflushed completion-events that can stay in the queue
// before flush kicks in.
maxUnflushedCompletionEvents =
conf.getInt(
MRJobConfig.MR_AM_HISTORY_MAX_UNFLUSHED_COMPLETE_EVENTS,
MRJobConfig.DEFAULT_MR_AM_HISTORY_MAX_UNFLUSHED_COMPLETE_EVENTS);
// We want to cut down flushes after job completes so as to write quicker,
// so we increase maxUnflushedEvents post Job completion by using the
// following multiplier.
postJobCompletionMultiplier =
conf.getInt(
MRJobConfig.MR_AM_HISTORY_JOB_COMPLETE_UNFLUSHED_MULTIPLIER,
MRJobConfig.DEFAULT_MR_AM_HISTORY_JOB_COMPLETE_UNFLUSHED_MULTIPLIER);
// Max time until which flush doesn't take place.
flushTimeout =
conf.getLong(
MRJobConfig.MR_AM_HISTORY_COMPLETE_EVENT_FLUSH_TIMEOUT_MS,
MRJobConfig.DEFAULT_MR_AM_HISTORY_COMPLETE_EVENT_FLUSH_TIMEOUT_MS);
minQueueSizeForBatchingFlushes =
conf.getInt(
MRJobConfig.MR_AM_HISTORY_USE_BATCHED_FLUSH_QUEUE_SIZE_THRESHOLD,
MRJobConfig.DEFAULT_MR_AM_HISTORY_USE_BATCHED_FLUSH_QUEUE_SIZE_THRESHOLD);
super.init(conf);
}
