Author: cutting
Date: Fri Dec 15 14:11:46 2006
New Revision: 487691
URL: http://svn.apache.org/viewvc?view=rev&rev=487691
Log:
HADOOP-791. Fix a deadlock in the task tracker. Contributed by Mahadev.
Modified:
lucene/hadoop/trunk/CHANGES.txt
lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/TaskTracker.java
Modified: lucene/hadoop/trunk/CHANGES.txt
URL:
http://svn.apache.org/viewvc/lucene/hadoop/trunk/CHANGES.txt?view=diff&rev=487691&r1=487690&r2=487691
==============================================================================
--- lucene/hadoop/trunk/CHANGES.txt (original)
+++ lucene/hadoop/trunk/CHANGES.txt Fri Dec 15 14:11:46 2006
@@ -110,6 +110,9 @@
2. HADOOP-827. Turn off speculative execution by default, since it's
currently broken. (omalley via cutting)
+ 3. HADOOP-791. Fix a deadlock in the task tracker.
+ (Mahadev Konar via cutting)
+
Release 0.9.1 - 2006-12-06
Modified: lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/TaskTracker.java
URL:
http://svn.apache.org/viewvc/lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/TaskTracker.java?view=diff&rev=487691&r1=487690&r2=487691
==============================================================================
--- lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/TaskTracker.java
(original)
+++ lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/TaskTracker.java Fri
Dec 15 14:11:46 2006
@@ -1130,28 +1130,33 @@
* We no longer need anything from this task, as the job has
* finished. If the task is still running, kill it (and clean up
*/
- public synchronized void jobHasFinished() throws IOException {
-
- if (getRunState() == TaskStatus.State.RUNNING) {
+ public void jobHasFinished() throws IOException {
+ boolean killTask = false;
+ synchronized(this){
+ killTask = (getRunState() == TaskStatus.State.RUNNING);
+ if (killTask) {
killAndCleanup(false);
- } else {
- cleanup();
- }
- if (keepJobFiles)
- return;
-
- // Delete temp directory in case any task used PhasedFileSystem.
- try{
- String systemDir = task.getConf().get("mapred.system.dir");
- Path taskTempDir = new Path(systemDir + "/" +
- task.getJobId() + "/" + task.getTipId());
- if( fs.exists(taskTempDir)){
- fs.delete(taskTempDir) ;
}
- }catch(IOException e){
- LOG.warn("Error in deleting reduce temporary output",e);
+ }
+ if (!killTask) {
+ cleanup();
+ }
+ if (keepJobFiles)
+ return;
+
+ synchronized(this){
+ // Delete temp directory in case any task used PhasedFileSystem.
+ try{
+ String systemDir = task.getConf().get("mapred.system.dir");
+ Path taskTempDir = new Path(systemDir + "/" +
+ task.getJobId() + "/" + task.getTipId() + "/" +
task.getTaskId());
+ if( fs.exists(taskTempDir)){
+ fs.delete(taskTempDir) ;
+ }
+ }catch(IOException e){
+ LOG.warn("Error in deleting reduce temporary output",e);
+ }
}
-
// Delete the job directory for this
// task if the job is done/failed
if (purgeJobFiles) {
@@ -1205,6 +1210,9 @@
* We no longer need anything from this task. Either the
* controlling job is all done and the files have been copied
* away, or the task failed and we don't need the remains.
+ * Any calls to cleanup should not lock the tip first.
+ * cleanup does the right thing- updates tasks in Tasktracker
+ * by locking tasktracker first and then locks the tip.
*/
void cleanup() throws IOException {
String taskId = task.getTaskId();
