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

YARN-189. Fixed a deadlock between RM's ApplicationMasterService and the dispatcher. Contributed by Thomas Graves.

Browse Source 
git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1404431 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Vinod Kumar Vavilapalli 2012-11-01 00:32:37 +00:00
parent 50993e9d48
commit a124297cf0
2 changed files with 21 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -196,6 +196,9 @@ Release 0.23.5 - UNRELEASED
YARN-166. capacity scheduler doesn't allow capacity < 1.0 (tgraves via
bobby)
YARN-189. Fixed a deadlock between RM's ApplicationMasterService and the
dispatcher. (Thomas Graves via vinodkv)
Release 0.23.4 - UNRELEASED
INCOMPATIBLE CHANGES

29
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ApplicationMasterService.java
View File

@ -265,10 +265,10 @@ public AllocateResponse allocate(AllocateRequest request)
// Oh damn! Sending reboot isn't enough. RM state is corrupted. TODO:
allocateResponse.setAMResponse(reboot);
return allocateResponse;
}
}
// Allow only one thread in AM to do heartbeat at a time.
synchronized (lastResponse) { // BUG TODO: Locking order is screwed.
synchronized (lastResponse) {
// Send the status update to the appAttempt.
this.rmContext.getDispatcher().getEventHandler().handle(
@ -282,7 +282,8 @@ public AllocateResponse allocate(AllocateRequest request)
Allocation allocation =
this.rScheduler.allocate(appAttemptId, ask, release);
RMApp app = this.rmContext.getRMApps().get(appAttemptId.getApplicationId());
RMApp app = this.rmContext.getRMApps().get(
appAttemptId.getApplicationId());
RMAppAttempt appAttempt = app.getRMAppAttempt(appAttemptId);
AMResponse response = recordFactory.newRecordInstance(AMResponse.class);
@ -316,7 +317,18 @@ public AllocateResponse allocate(AllocateRequest request)
.pullJustFinishedContainers());
response.setResponseId(lastResponse.getResponseId() + 1);
response.setAvailableResources(allocation.getResourceLimit());
responseMap.put(appAttemptId, response);
AMResponse oldResponse = responseMap.put(appAttemptId, response);
if (oldResponse == null) {
// appAttempt got unregistered, remove it back out
responseMap.remove(appAttemptId);
String message = "App Attempt removed from the cache during allocate"
+ appAttemptId;
LOG.error(message);
allocateResponse.setAMResponse(reboot);
return allocateResponse;
}
allocateResponse.setAMResponse(response);
allocateResponse.setNumClusterNodes(this.rScheduler.getNumClusterNodes());
return allocateResponse;
@ -331,12 +343,7 @@ public void registerAppAttempt(ApplicationAttemptId attemptId) {
}
public void unregisterAttempt(ApplicationAttemptId attemptId) {
AMResponse lastResponse = responseMap.get(attemptId);
if (lastResponse != null) {
synchronized (lastResponse) {
responseMap.remove(attemptId);
}
}
responseMap.remove(attemptId);
}
public void refreshServiceAcls(Configuration configuration,