[TEZ-1929] pre-empted tasks should be marked as killed instead of failed

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.6.0
• Fix Version/s: 0.6.1
• Target Version/s: 0.6.1
• Hadoop Flags: Reviewed
• Assignee: Bikas Saha
• Reporter: Rajesh Balamohan
• Created: 08/Jan/15 08:57
• Updated: 30/Jun/15 04:52
• Resolved: 31/Jan/15 00:40

Description

Observed this behavior 3 or 4 times

• Ran a hive query with tez (query_17 at 10 TB scale)
• Occasionally, Map_7 task will get into failed state in the middle of fetching data from other sources (only one task is available in Map_7).

2015-01-08 00:19:10,289 INFO [Fetcher [Map_6] #0] impl.ShuffleManager: Completed fetch for attempt: InputAttemptIdentifier [inputIdentifier=InputIdentifier [inputIndex=0], attemptNumber=0, pathComponent=attempt_1420000126204_0233_1_06_000000_0_10003] to MEMORY, CompressedSize=6757, DecompressedSize=16490,EndTime=1420705150289, TimeTaken=5, Rate=1.29 MB/s
2015-01-08 00:19:10,290 INFO [Fetcher [Map_6] #0] impl.ShuffleManager: All inputs fetched for input vertex : Map 6
2015-01-08 00:19:10,290 INFO [Fetcher [Map_6] #0] impl.ShuffleManager: copy(0 of 1. Transfer rate (CumulativeDataFetched/TimeSinceInputStarted)) 0.01 MB/s)
2015-01-08 00:19:10,290 INFO [ShuffleRunner [Map_6]] impl.ShuffleManager: Shutting down FetchScheduler, Was Interrupted: false
2015-01-08 00:19:10,290 INFO [ShuffleRunner [Map_6]] impl.ShuffleManager: Scheduler thread completed
2015-01-08 00:19:41,986 INFO [TaskHeartbeatThread] task.TaskReporter: Received should die response from AM
2015-01-08 00:19:41,986 INFO [TaskHeartbeatThread] task.TaskReporter: Asked to die via task heartbeat
2015-01-08 00:19:41,987 INFO [main] task.TezTaskRunner: Interrupted while waiting for task to complete. Interrupting task
2015-01-08 00:19:41,987 INFO [main] task.TezTaskRunner: Shutdown requested... returning
2015-01-08 00:19:41,987 INFO [main] task.TezChild: Got a shouldDie notification via hearbeats. Shutting down
2015-01-08 00:19:41,990 ERROR [TezChild] tez.TezProcessor: java.lang.InterruptedException
	at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.reportInterruptAfterWait(AbstractQueuedSynchronizer.java:2014)
	at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2048)
	at org.apache.tez.runtime.InputReadyTracker$InputReadyMonitor.awaitCondition(InputReadyTracker.java:120)
	at org.apache.tez.runtime.InputReadyTracker.waitForAnyInputReady(InputReadyTracker.java:83)
	at org.apache.tez.runtime.api.impl.TezProcessorContextImpl.waitForAnyInputReady(TezProcessorContextImpl.java:106)
	at org.apache.hadoop.hive.ql.exec.tez.TezProcessor.initializeAndRunProcessor(TezProcessor.java:153)
	at org.apache.hadoop.hive.ql.exec.tez.TezProcessor.run(TezProcessor.java:138)
	at org.apache.tez.runtime.LogicalIOProcessorRuntimeTask.run(LogicalIOProcessorRuntimeTask.java:328)
	at org.apache.tez.runtime.task.TezTaskRunner$TaskRunnerCallable$1.run(TezTaskRunner.java:179)
	at org.apache.tez.runtime.task.TezTaskRunner$TaskRunnerCallable$1.run(TezTaskRunner.java:171)

From the initial look, it appears that TaskAttemptListenerImpTezDag.heartbeat is unable to identify the containerId from registeredContainers. Need to verify this.

I will attach the sample task log and the tez-ui details.