Repository: hadoop Updated Branches: refs/heads/branch-2.6 b2d02863a -> 74027c24c
YARN-4546. ResourceManager crash due to scheduling opportunity overflow. Contributed by Jason Lowe. (cherry picked from commit c1462a67ff7bb632df50e1c52de971cced56c6a3) (cherry picked from commit 1cc001db4c3767072b5d065d161bc5c6d1c480d4) Conflicts: hadoop-yarn-project/CHANGES.txt hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java Project: http://git-wip-us.apache.org/repos/asf/hadoop/repo Commit: http://git-wip-us.apache.org/repos/asf/hadoop/commit/c0ffe25a Tree: http://git-wip-us.apache.org/repos/asf/hadoop/tree/c0ffe25a Diff: http://git-wip-us.apache.org/repos/asf/hadoop/diff/c0ffe25a Branch: refs/heads/branch-2.6 Commit: c0ffe25a6547dbb069993f87e20d9c723c4686dc Parents: b2d0286 Author: Junping Du <junping...@apache.org> Authored: Wed Jan 6 05:49:24 2016 -0800 Committer: Junping Du <junping...@apache.org> Committed: Wed Jan 6 05:56:16 2016 -0800 ---------------------------------------------------------------------- hadoop-yarn-project/CHANGES.txt | 3 +++ .../scheduler/SchedulerApplicationAttempt.java | 11 ++++++++-- .../TestSchedulerApplicationAttempt.java | 22 ++++++++++++++++++++ 3 files changed, 34 insertions(+), 2 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/hadoop/blob/c0ffe25a/hadoop-yarn-project/CHANGES.txt ---------------------------------------------------------------------- diff --git a/hadoop-yarn-project/CHANGES.txt b/hadoop-yarn-project/CHANGES.txt index 38089a1..6675dce 100644 --- a/hadoop-yarn-project/CHANGES.txt +++ b/hadoop-yarn-project/CHANGES.txt @@ -28,6 +28,9 @@ Release 2.6.4 - UNRELEASED YARN-3697. FairScheduler: ContinuousSchedulingThread can fail to shutdown. (Zhihai Xu via kasha) + YARN-4546. ResourceManager crash due to scheduling opportunity overflow. + (Jason Lowe via junping_du) + Release 2.6.3 - 2015-12-17 INCOMPATIBLE CHANGES http://git-wip-us.apache.org/repos/asf/hadoop/blob/c0ffe25a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java ---------------------------------------------------------------------- diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java index 510276c..5629734 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/SchedulerApplicationAttempt.java @@ -482,8 +482,10 @@ public class SchedulerApplicationAttempt { } public synchronized void addSchedulingOpportunity(Priority priority) { - schedulingOpportunities.setCount(priority, - schedulingOpportunities.count(priority) + 1); + int count = schedulingOpportunities.count(priority); + if (count < Integer.MAX_VALUE) { + schedulingOpportunities.setCount(priority, count + 1); + } } public synchronized void subtractSchedulingOpportunity(Priority priority) { @@ -517,6 +519,11 @@ public class SchedulerApplicationAttempt { schedulingOpportunities.setCount(priority, 0); } + @VisibleForTesting + void setSchedulingOpportunities(Priority priority, int count) { + schedulingOpportunities.setCount(priority, count); + } + synchronized AggregateAppResourceUsage getRunningAggregateAppResourceUsage() { long currentTimeMillis = System.currentTimeMillis(); // Don't walk the whole container list if the resources were computed http://git-wip-us.apache.org/repos/asf/hadoop/blob/c0ffe25a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java ---------------------------------------------------------------------- diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java index c648b83..973c0b6 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/TestSchedulerApplicationAttempt.java @@ -163,4 +163,26 @@ public class TestSchedulerApplicationAttempt { ApplicationAttemptId.newInstance(appIdImpl, attemptId); return attId; } + + @Test + public void testSchedulingOpportunityOverflow() throws Exception { + ApplicationAttemptId attemptId = createAppAttemptId(0, 0); + Queue queue = createQueue("test", null); + RMContext rmContext = mock(RMContext.class); + when(rmContext.getEpoch()).thenReturn(3L); + SchedulerApplicationAttempt app = new SchedulerApplicationAttempt( + attemptId, "user", queue, queue.getActiveUsersManager(), rmContext); + Priority priority = Priority.newInstance(1); + assertEquals(0, app.getSchedulingOpportunities(priority)); + app.addSchedulingOpportunity(priority); + assertEquals(1, app.getSchedulingOpportunities(priority)); + // verify the count is capped at MAX_VALUE and does not overflow + app.setSchedulingOpportunities(priority, Integer.MAX_VALUE - 1); + assertEquals(Integer.MAX_VALUE - 1, + app.getSchedulingOpportunities(priority)); + app.addSchedulingOpportunity(priority); + assertEquals(Integer.MAX_VALUE, app.getSchedulingOpportunities(priority)); + app.addSchedulingOpportunity(priority); + assertEquals(Integer.MAX_VALUE, app.getSchedulingOpportunities(priority)); + } }