[MAPREDUCE-7353] Mapreduce job fails when NM is stopped

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.4.0, 2.10.2, 3.2.3, 3.3.2
• Fix Version/s: 3.4.0, 2.10.2, 3.2.3, 3.3.2
• Component/s: task
• Target Version/s: 3.4.0, 2.10.2, 3.2.3, 3.3.2
• Hadoop Flags: Reviewed
• Assignee: Bilwa S T
• Reporter: Bilwa S T
• Created: 16/Jun/21 07:13
• Updated: 11/Feb/24 14:21
• Resolved: 07/Jul/21 21:59

Description

Job fails as task fail due to too many fetch failures 

Line 48048: 2021-06-02 16:25:02,002 | INFO  | ContainerLauncher #6 | Processing the event EventType: CONTAINER_REMOTE_CLEANUP for container container_e03_1622107691213_1054_01_000005 taskAttempt attempt_1622107691213_1054_m_000000_0 | ContainerLauncherImpl.java:394
	Line 48053: 2021-06-02 16:25:02,002 | INFO  | ContainerLauncher #6 | KILLING attempt_1622107691213_1054_m_000000_0 | ContainerLauncherImpl.java:209
	Line 58026: 2021-06-02 16:26:34,034 | INFO  | AsyncDispatcher event handler | TaskAttempt killed because it ran on unusable node node-group-1ZYEq0002:26009. AttemptId:attempt_1622107691213_1054_m_000000_0 | JobImpl.java:1401
	Line 58030: 2021-06-02 16:26:34,034 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_KILL | TaskAttemptImpl.java:1390
	Line 58035: 2021-06-02 16:26:34,034 | INFO  | RMCommunicator Allocator | Killing taskAttempt:attempt_1622107691213_1054_m_000000_0 because it is running on unusable node:node-group-1ZYEq0002:26009 | RMContainerAllocator.java:1066
	Line 58043: 2021-06-02 16:26:34,034 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_KILL | TaskAttemptImpl.java:1390
	Line 58054: 2021-06-02 16:26:34,034 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_DIAGNOSTICS_UPDATE | TaskAttemptImpl.java:1390
	Line 58055: 2021-06-02 16:26:34,034 | INFO  | AsyncDispatcher event handler | Diagnostics report from attempt_1622107691213_1054_m_000000_0: Container released on a *lost* node | TaskAttemptImpl.java:2649
	Line 58057: 2021-06-02 16:26:34,034 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_KILL | TaskAttemptImpl.java:1390
	Line 60317: 2021-06-02 16:26:57,057 | INFO  | AsyncDispatcher event handler | Too many fetch-failures for output of task attempt: attempt_1622107691213_1054_m_000000_0 ... raising fetch failure to map | JobImpl.java:2005
	Line 60319: 2021-06-02 16:26:57,057 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_TOO_MANY_FETCH_FAILURE | TaskAttemptImpl.java:1390
	Line 60320: 2021-06-02 16:26:57,057 | INFO  | AsyncDispatcher event handler | attempt_1622107691213_1054_m_000000_0 transitioned from state SUCCESS_CONTAINER_CLEANUP to FAILED, event type is TA_TOO_MANY_FETCH_FAILURE and nodeId=node-group-1ZYEq0002:26009 | TaskAttemptImpl.java:1411
	Line 69487: 2021-06-02 16:30:02,002 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_DIAGNOSTICS_UPDATE | TaskAttemptImpl.java:1390
	Line 69527: 2021-06-02 16:30:02,002 | INFO  | AsyncDispatcher event handler | Diagnostics report from attempt_1622107691213_1054_m_000000_0: cleanup failed for container container_e03_1622107691213_1054_01_000005 : java.net.ConnectException: Call From node-group-1ZYEq0001/192.168.0.66 to node-group-1ZYEq0002:26009 failed on connection exception: java.net.ConnectException: Connection refused; For more details see:  http://wiki.apache.org/hadoop/ConnectionRefused
	Line 69607: 2021-06-02 16:30:02,002 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_CONTAINER_CLEANED | TaskAttemptImpl.java:1390
	Line 69609: 2021-06-02 16:30:02,002 | DEBUG | AsyncDispatcher event handler | Processing attempt_1622107691213_1054_m_000000_0 of type TA_CONTAINER_CLEANED | TaskAttemptImpl.java:1390
	Line 73645: 2021-06-02 16:23:56,056 | DEBUG | fetcher#9 | Fetcher 9 going to fetch from node-group-1ZYEq0002:26008 for: [attempt_1622107691213_1054_m_000000_0] | Fetcher.java:318
	Line 73646: 2021-06-02 16:23:56,056 | DEBUG | fetcher#9 | MapOutput URL for node-group-1ZYEq0002:26008 -> http://node-group-1ZYEq0002:26008/mapOutput?job=job_1622107691213_1054&reduce=4&map=attempt_1622107691213_1054_m_000000_0 | Fetcher.java:686
	Line 74093: 2021-06-02 16:26:56,056 | INFO  | fetcher#9 | Reporting fetch failure for attempt_1622107691213_1054_m_000000_0 to MRAppMaster. | ShuffleSchedulerImpl.java:349

As we can see from logs that RM reported AM about node update at 16:26:34 but event was skipped as KILL event is ignored when TaskAttemptImpl is in SUCCESS_CONTAINER_CLEANUP state. So next we receive TA_TOO_MANY_FETCH_FAILURE event which will lead to task fail.