Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-7999. FsDatasetImpl#createTemporary sometimes holds the FSDatasetImpl lock for a very long time (sinago via cmccabe) 

(cherry picked from commit 28bebc81db)
This commit is contained in:
Colin Patrick Mccabe 2015-04-06 08:54:46 -07:00
parent d2b2d76cce
commit a827089905
2 changed files with 52 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
View File

@ -1078,6 +1078,9 @@ Release 2.7.0 - UNRELEASED
HDFS-8051. FsVolumeList#addVolume should release volume reference if not HDFS-8051. FsVolumeList#addVolume should release volume reference if not
put it into BlockScanner. (Lei (Eddy) Xu via Colin P. McCabe) put it into BlockScanner. (Lei (Eddy) Xu via Colin P. McCabe)
HDFS-7999. FsDatasetImpl#createTemporary sometimes holds the FSDatasetImpl
lock for a very long time (sinago via cmccabe)
BREAKDOWN OF HDFS-7584 SUBTASKS AND RELATED JIRAS BREAKDOWN OF HDFS-7584 SUBTASKS AND RELATED JIRAS
HDFS-7720. Quota by Storage Type API, tools and ClientNameNode HDFS-7720. Quota by Storage Type API, tools and ClientNameNode

59
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java
View File

@ -1414,24 +1414,21 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
} }
@Override // FsDatasetSpi @Override // FsDatasetSpi
public synchronized ReplicaHandler createTemporary( public ReplicaHandler createTemporary(
StorageType storageType, ExtendedBlock b) throws IOException { StorageType storageType, ExtendedBlock b) throws IOException {
ReplicaInfo replicaInfo = volumeMap.get(b.getBlockPoolId(), b.getBlockId()); long startTimeMs = Time.monotonicNow();
if (replicaInfo != null) { long writerStopTimeoutMs = datanode.getDnConf().getXceiverStopTimeout();
if (replicaInfo.getGenerationStamp() < b.getGenerationStamp() ReplicaInfo lastFoundReplicaInfo = null;
&& replicaInfo instanceof ReplicaInPipeline) { do {
// Stop the previous writer synchronized (this) {
((ReplicaInPipeline)replicaInfo) ReplicaInfo currentReplicaInfo =
.stopWriter(datanode.getDnConf().getXceiverStopTimeout()); volumeMap.get(b.getBlockPoolId(), b.getBlockId());
invalidate(b.getBlockPoolId(), new Block[]{replicaInfo}); if (currentReplicaInfo == lastFoundReplicaInfo) {
} else { if (lastFoundReplicaInfo != null) {
throw new ReplicaAlreadyExistsException("Block " + b + invalidate(b.getBlockPoolId(), new Block[] { lastFoundReplicaInfo });
" already exists in state " + replicaInfo.getState() +
" and thus cannot be created.");
} }
} FsVolumeReference ref =
volumes.getNextVolume(storageType, b.getNumBytes());
FsVolumeReference ref = volumes.getNextVolume(storageType, b.getNumBytes());
FsVolumeImpl v = (FsVolumeImpl) ref.getVolume(); FsVolumeImpl v = (FsVolumeImpl) ref.getVolume();
// create a temporary file to hold block in the designated volume // create a temporary file to hold block in the designated volume
File f; File f;
@ -1441,11 +1438,35 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
IOUtils.cleanup(null, ref); IOUtils.cleanup(null, ref);
throw e; throw e;
} }
ReplicaInPipeline newReplicaInfo =
ReplicaInPipeline newReplicaInfo = new ReplicaInPipeline(b.getBlockId(), new ReplicaInPipeline(b.getBlockId(), b.getGenerationStamp(), v,
b.getGenerationStamp(), v, f.getParentFile(), 0); f.getParentFile(), 0);
volumeMap.add(b.getBlockPoolId(), newReplicaInfo); volumeMap.add(b.getBlockPoolId(), newReplicaInfo);
return new ReplicaHandler(newReplicaInfo, ref); return new ReplicaHandler(newReplicaInfo, ref);
} else {
if (!(currentReplicaInfo.getGenerationStamp() < b
.getGenerationStamp() && currentReplicaInfo instanceof ReplicaInPipeline)) {
throw new ReplicaAlreadyExistsException("Block " + b
+ " already exists in state " + currentReplicaInfo.getState()
+ " and thus cannot be created.");
}
lastFoundReplicaInfo = currentReplicaInfo;
}
}
// Hang too long, just bail out. This is not supposed to happen.
long writerStopMs = Time.monotonicNow() - startTimeMs;
if (writerStopMs > writerStopTimeoutMs) {
LOG.warn("Unable to stop existing writer for block " + b + " after "
+ writerStopMs + " miniseconds.");
throw new IOException("Unable to stop existing writer for block " + b
+ " after " + writerStopMs + " miniseconds.");
}
// Stop the previous writer
((ReplicaInPipeline) lastFoundReplicaInfo)
.stopWriter(writerStopTimeoutMs);
} while (true);
} }
/** /**