YARN-3675. FairScheduler: RM quits when node removal races with continuous-scheduling on the same node. (Anubhav Dhoot via kasha)
This commit is contained in:
parent
05e04f34f2
commit
4513761869
@ -541,6 +541,9 @@ Release 2.7.1 - UNRELEASED
|
||||
YARN-3646. Applications are getting stuck some times in case of retry
|
||||
policy forever. (Raju Bairishetti via devaraj)
|
||||
|
||||
YARN-3675. FairScheduler: RM quits when node removal races with
|
||||
continuous-scheduling on the same node. (Anubhav Dhoot via kasha)
|
||||
|
||||
Release 2.7.0 - 2015-04-20
|
||||
|
||||
INCOMPATIBLE CHANGES
|
||||
|
@ -1039,13 +1039,23 @@ public int compare(NodeId n1, NodeId n2) {
|
||||
nodes.get(n1).getAvailableResource());
|
||||
}
|
||||
}
|
||||
|
||||
private synchronized void attemptScheduling(FSSchedulerNode node) {
|
||||
|
||||
@VisibleForTesting
|
||||
synchronized void attemptScheduling(FSSchedulerNode node) {
|
||||
if (rmContext.isWorkPreservingRecoveryEnabled()
|
||||
&& !rmContext.isSchedulerReadyForAllocatingContainers()) {
|
||||
return;
|
||||
}
|
||||
|
||||
final NodeId nodeID = node.getNodeID();
|
||||
if (!nodes.containsKey(nodeID)) {
|
||||
// The node might have just been removed while this thread was waiting
|
||||
// on the synchronized lock before it entered this synchronized method
|
||||
LOG.info("Skipping scheduling as the node " + nodeID +
|
||||
" has been removed");
|
||||
return;
|
||||
}
|
||||
|
||||
// Assign new containers...
|
||||
// 1. Check for reserved applications
|
||||
// 2. Schedule if there are no reservations
|
||||
|
@ -3889,6 +3889,50 @@ public void testContinuousSchedulingWithNodeRemoved() throws Exception {
|
||||
}
|
||||
}
|
||||
|
||||
@Test
|
||||
public void testSchedulingOnRemovedNode() throws Exception {
|
||||
// Disable continuous scheduling, will invoke continuous scheduling manually
|
||||
scheduler.init(conf);
|
||||
scheduler.start();
|
||||
Assert.assertTrue("Continuous scheduling should be disabled.",
|
||||
!scheduler.isContinuousSchedulingEnabled());
|
||||
|
||||
ApplicationAttemptId id11 = createAppAttemptId(1, 1);
|
||||
createMockRMApp(id11);
|
||||
|
||||
scheduler.addApplication(id11.getApplicationId(), "root.queue1", "user1",
|
||||
false);
|
||||
scheduler.addApplicationAttempt(id11, false, false);
|
||||
|
||||
List<ResourceRequest> ask1 = new ArrayList<>();
|
||||
ResourceRequest request1 =
|
||||
createResourceRequest(1024, 8, ResourceRequest.ANY, 1, 1, true);
|
||||
|
||||
ask1.add(request1);
|
||||
scheduler.allocate(id11, ask1, new ArrayList<ContainerId>(), null,
|
||||
null);
|
||||
|
||||
String hostName = "127.0.0.1";
|
||||
RMNode node1 = MockNodes.newNodeInfo(1,
|
||||
Resources.createResource(8 * 1024, 8), 1, hostName);
|
||||
NodeAddedSchedulerEvent nodeEvent1 = new NodeAddedSchedulerEvent(node1);
|
||||
scheduler.handle(nodeEvent1);
|
||||
|
||||
FSSchedulerNode node = (FSSchedulerNode)scheduler.getSchedulerNode(
|
||||
node1.getNodeID());
|
||||
|
||||
NodeRemovedSchedulerEvent removeNode1 =
|
||||
new NodeRemovedSchedulerEvent(node1);
|
||||
scheduler.handle(removeNode1);
|
||||
|
||||
scheduler.attemptScheduling(node);
|
||||
|
||||
AppAttemptRemovedSchedulerEvent appRemovedEvent1 =
|
||||
new AppAttemptRemovedSchedulerEvent(id11,
|
||||
RMAppAttemptState.FINISHED, false);
|
||||
scheduler.handle(appRemovedEvent1);
|
||||
}
|
||||
|
||||
@Test
|
||||
public void testDefaultRuleInitializesProperlyWhenPolicyNotConfigured()
|
||||
throws IOException {
|
||||
|
Loading…
Reference in New Issue
Block a user