big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

MAPREDUCE-3721. Fixed a race in shuffle which caused reduces to hang. Contributed by Siddharth Seth.

Browse Source 
git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1236041 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Arun Murthy 2012-01-26 05:40:29 +00:00
parent 21c9116309
commit fae75c2d7f
2 changed files with 29 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -568,6 +568,9 @@ Release 0.23.1 - Unreleased
CapacityScheduler so that it deducts current-usage per user and not
per-application. (Arun C Murthy via vinodkv)
MAPREDUCE-3721. Fixed a race in shuffle which caused reduces to hang.
(sseth via acmurthy)
Release 0.23.0 - 2011-11-01
INCOMPATIBLE CHANGES

38
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/task/reduce/MergeManager.java
View File

@ -92,6 +92,7 @@ public class MergeManager<K, V> {
private final long memoryLimit;
private long usedMemory;
private long commitMemory;
private final long maxSingleShuffleLimit;
private final int memToMemMergeOutputsThreshold;
@ -181,6 +182,13 @@ public MergeManager(TaskAttemptID reduceId, JobConf jobConf,
"ioSortFactor=" + ioSortFactor + ", " +
"memToMemMergeOutputsThreshold=" + memToMemMergeOutputsThreshold);
if (this.maxSingleShuffleLimit >= this.mergeThreshold) {
throw new RuntimeException("Invlaid configuration: "
+ "maxSingleShuffleLimit should be less than mergeThreshold"
+ "maxSingleShuffleLimit: " + this.maxSingleShuffleLimit
+ "mergeThreshold: " + this.mergeThreshold);
}
boolean allowMemToMemMerge =
jobConf.getBoolean(MRJobConfig.REDUCE_MEMTOMEM_ENABLED, false);
if (allowMemToMemMerge) {
@ -245,16 +253,16 @@ public synchronized MapOutput<K,V> reserve(TaskAttemptID mapId,
// all the stalled threads
if (usedMemory > memoryLimit) {
LOG.debug(mapId + ": Stalling shuffle since usedMemory (" + usedMemory +
") is greater than memoryLimit (" + memoryLimit + ")");
LOG.debug(mapId + ": Stalling shuffle since usedMemory (" + usedMemory
+ ") is greater than memoryLimit (" + memoryLimit + ")." +
" CommitMemory is (" + commitMemory + ")");
return stallShuffle;
}
// Allow the in-memory shuffle to progress
LOG.debug(mapId + ": Proceeding with shuffle since usedMemory (" +
usedMemory +
") is lesser than memoryLimit (" + memoryLimit + ")");
LOG.debug(mapId + ": Proceeding with shuffle since usedMemory ("
+ usedMemory + ") is lesser than memoryLimit (" + memoryLimit + ")."
+ "CommitMemory is (" + commitMemory + ")");
return unconditionalReserve(mapId, requestedSize, true);
}
@ -270,18 +278,24 @@ private synchronized MapOutput<K, V> unconditionalReserve(
}
synchronized void unreserve(long size) {
commitMemory -= size;
usedMemory -= size;
}
public synchronized void closeInMemoryFile(MapOutput<K,V> mapOutput) {
inMemoryMapOutputs.add(mapOutput);
LOG.info("closeInMemoryFile -> map-output of size: " + mapOutput.getSize()
+ ", inMemoryMapOutputs.size() -> " + inMemoryMapOutputs.size());
+ ", inMemoryMapOutputs.size() -> " + inMemoryMapOutputs.size()
+ ", commitMemory -> " + commitMemory + ", usedMemory ->" + usedMemory);
commitMemory+= mapOutput.getSize();
synchronized (inMemoryMerger) {
if (!inMemoryMerger.isInProgress() && usedMemory >= mergeThreshold) {
LOG.info("Starting inMemoryMerger's merge since usedMemory=" +
usedMemory + " > mergeThreshold=" + mergeThreshold);
// Can hang if mergeThreshold is really low.
if (!inMemoryMerger.isInProgress() && commitMemory >= mergeThreshold) {
LOG.info("Starting inMemoryMerger's merge since commitMemory=" +
commitMemory + " > mergeThreshold=" + mergeThreshold +
". Current usedMemory=" + usedMemory);
inMemoryMapOutputs.addAll(inMemoryMergedMapOutputs);
inMemoryMergedMapOutputs.clear();
inMemoryMerger.startMerge(inMemoryMapOutputs);