Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
None
-
None
-
None
-
None
Description
Starting from hadoop-2.5, we could know whether container is failed due to exceed allocated physical memory from its ContainerExitStatus. We could allocate more memory for the next TaskAttempt. But it may break the build with hadoop before 2.5, Need to think of ways to support the old hadoop version without this feature.