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

YARN-4794. Deadlock in NMClientImpl. Contributed by Jian He.

Browse Source
This commit is contained in:
Rohith Sharma K S 2016-04-12 09:59:33 +05:30
parent 44bbc50d91
commit ff722bbbdf
1 changed files with 7 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-client/src/main/java/org/apache/hadoop/yarn/client/api/impl/NMClientImpl.java
View File

@ -171,8 +171,6 @@ private void addStartingContainer(StartedContainer startedContainer)
throw RPCUtil.getRemoteException("Container "
+ startedContainer.containerId.toString() + " is already started");
}
startedContainers
.put(startedContainer.getContainerId(), startedContainer);
}
@Override
@ -182,7 +180,8 @@ public Map<String, ByteBuffer> startContainer(
// Do synchronization on StartedContainer to prevent race condition
// between startContainer and stopContainer only when startContainer is
// in progress for a given container.
StartedContainer startingContainer = createStartedContainer(container);
StartedContainer startingContainer =
new StartedContainer(container.getId(), container.getNodeId());
synchronized (startingContainer) {
addStartingContainer(startingContainer);
@ -210,18 +209,14 @@ public Map<String, ByteBuffer> startContainer(
}
allServiceResponse = response.getAllServicesMetaData();
startingContainer.state = ContainerState.RUNNING;
} catch (YarnException e) {
} catch (YarnException | IOException e) {
startingContainer.state = ContainerState.COMPLETE;
// Remove the started container if it failed to start
removeStartedContainer(startingContainer);
throw e;
} catch (IOException e) {
startingContainer.state = ContainerState.COMPLETE;
removeStartedContainer(startingContainer);
startedContainers.remove(startingContainer.containerId);
throw e;
} catch (Throwable t) {
startingContainer.state = ContainerState.COMPLETE;
removeStartedContainer(startingContainer);
startedContainers.remove(startingContainer.containerId);
throw RPCUtil.getRemoteException(t);
} finally {
if (proxy != null) {
@ -263,7 +258,7 @@ public void increaseContainerResource(Container container)
@Override
public void stopContainer(ContainerId containerId, NodeId nodeId)
throws YarnException, IOException {
StartedContainer startedContainer = getStartedContainer(containerId);
StartedContainer startedContainer = startedContainers.get(containerId);
// Only allow one request of stopping the container to move forward
// When entering the block, check whether the precursor has already stopped
@ -276,7 +271,7 @@ public void stopContainer(ContainerId containerId, NodeId nodeId)
stopContainerInternal(containerId, nodeId);
// Only after successful
startedContainer.state = ContainerState.COMPLETE;
removeStartedContainer(startedContainer);
startedContainers.remove(startedContainer.containerId);
}
} else {
stopContainerInternal(containerId, nodeId);
@ -333,23 +328,6 @@ private void stopContainerInternal(ContainerId containerId, NodeId nodeId)
}
}
}
protected synchronized StartedContainer createStartedContainer(
Container container) throws YarnException, IOException {
StartedContainer startedContainer = new StartedContainer(container.getId(),
container.getNodeId());
return startedContainer;
}
protected synchronized void
removeStartedContainer(StartedContainer container) {
startedContainers.remove(container.containerId);
}
protected synchronized StartedContainer getStartedContainer(
ContainerId containerId) {
return startedContainers.get(containerId);
}
public AtomicBoolean getCleanupRunningContainers() {
return cleanupRunningContainers;