YARN-6872. [Addendum patch] Ensure apps could run given NodeLabels are disabled post RM switchover/restart. Contributed by Sunil G
This commit is contained in:
parent
3a0b84d9c9
commit
f9139ac8f6
@ -519,20 +519,10 @@ private RMContainer recoverAndCreateContainer(NMContainerStatus status,
|
||||
container.setVersion(status.getVersion());
|
||||
ApplicationAttemptId attemptId =
|
||||
container.getId().getApplicationAttemptId();
|
||||
String labelExpression = status.getNodeLabelExpression();
|
||||
// If NodeLabel is disabled but recovered container has label expression
|
||||
// its better to suppress that and considered as default label.
|
||||
if (!status.getNodeLabelExpression().isEmpty() && !YarnConfiguration
|
||||
.areNodeLabelsEnabled(rmContext.getYarnConfiguration())) {
|
||||
labelExpression = RMNodeLabelsManager.NO_LABEL;
|
||||
}
|
||||
|
||||
RMContainer rmContainer =
|
||||
new RMContainerImpl(container,
|
||||
SchedulerRequestKey.extractFrom(container), attemptId,
|
||||
node.getNodeID(), applications.get(
|
||||
attemptId.getApplicationId()).getUser(), rmContext,
|
||||
status.getCreationTime(), labelExpression);
|
||||
RMContainer rmContainer = new RMContainerImpl(container,
|
||||
SchedulerRequestKey.extractFrom(container), attemptId, node.getNodeID(),
|
||||
applications.get(attemptId.getApplicationId()).getUser(), rmContext,
|
||||
status.getCreationTime(), status.getNodeLabelExpression());
|
||||
return rmContainer;
|
||||
}
|
||||
|
||||
|
@ -524,7 +524,7 @@ public void transferStateFromPreviousAppSchedulingInfo(
|
||||
this.placesBlacklistedByApp = appInfo.getBlackList();
|
||||
}
|
||||
|
||||
public void recoverContainer(RMContainer rmContainer) {
|
||||
public void recoverContainer(RMContainer rmContainer, String partition) {
|
||||
try {
|
||||
this.writeLock.lock();
|
||||
QueueMetrics metrics = queue.getMetrics();
|
||||
@ -540,8 +540,8 @@ public void recoverContainer(RMContainer rmContainer) {
|
||||
return;
|
||||
}
|
||||
|
||||
metrics.allocateResources(rmContainer.getNodeLabelExpression(),
|
||||
user, 1, rmContainer.getAllocatedResource(), false);
|
||||
metrics.allocateResources(partition, user, 1,
|
||||
rmContainer.getAllocatedResource(), false);
|
||||
} finally {
|
||||
this.writeLock.unlock();
|
||||
}
|
||||
|
@ -1103,7 +1103,7 @@ public void recoverContainer(SchedulerNode node,
|
||||
try {
|
||||
writeLock.lock();
|
||||
// recover app scheduling info
|
||||
appSchedulingInfo.recoverContainer(rmContainer);
|
||||
appSchedulingInfo.recoverContainer(rmContainer, node.getPartition());
|
||||
|
||||
if (rmContainer.getState().equals(RMContainerState.COMPLETED)) {
|
||||
return;
|
||||
|
Loading…
Reference in New Issue
Block a user