[HIVE-23068] Error when submitting fragment to LLAP via external client: IllegalStateException: Only a single registration allowed per entity

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0.0-alpha-1
• Component/s: llap
• Assignee: Jason Dere
• Reporter: Jason Dere
• Created: 23/Mar/20 22:43
• Updated: 17/Nov/22 08:51
• Resolved: 29/May/20 23:18

Description

LLAP external client (via hive-warehouse-connector) somehow seems to be sending duplicate submissions for the same fragment/attempt. When the 2nd request is sent this results in the following error:

2020-03-17T06:49:11,239 WARN  [IPC Server handler 2 on 15001 ()] org.apache.hadoop.ipc.Server: IPC Server handler 2 on 15001, call Call#75 Retry#0 org.apache.hadoop.hive.llap.protocol.LlapProtocolBlockingPB.submitWork from 19.40.252.114:33906
java.lang.IllegalStateException: Only a single registration allowed per entity. Duplicate for TaskWrapper{task=attempt_1854104024183112753_6052_0_00_000128_1, inWaitQueue=true, inPreemptionQueue=false, registeredForNotifications=true, canFinish=true, canFinish(in queue)=true, isGuaranteed=false, firstAttemptStartTime=1584442003327, dagStartTime=1584442003327, withinDagPriority=0, vertexParallelism= 2132, selfAndUpstreamParallelism= 2132, selfAndUpstreamComplete= 0}
    at org.apache.hadoop.hive.llap.daemon.impl.QueryInfo$FinishableStateTracker.registerForUpdates(QueryInfo.java:233) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.QueryInfo.registerForFinishableStateUpdates(QueryInfo.java:205) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.QueryFragmentInfo.registerForFinishableStateUpdates(QueryFragmentInfo.java:160) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService$TaskWrapper.maybeRegisterForFinishedStateNotifications(TaskExecutorService.java:1167) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService.schedule(TaskExecutorService.java:564) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService.schedule(TaskExecutorService.java:93) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.ContainerRunnerImpl.submitWork(ContainerRunnerImpl.java:292) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.LlapDaemon.submitWork(LlapDaemon.java:610) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.LlapProtocolServerImpl.submitWork(LlapProtocolServerImpl.java:122) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.rpc.LlapDaemonProtocolProtos$LlapDaemonProtocol$2.callBlockingMethod(LlapDaemonProtocolProtos.java:22695) ~[hive-exec-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.32-1]
    at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:524) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]
    at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:1025) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]
    at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:876) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]
    at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:822) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]
    at java.security.AccessController.doPrivileged(Native Method) ~[?:1.8.0_191]
    at javax.security.auth.Subject.doAs(Subject.java:422) ~[?:1.8.0_191]
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1730) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]
    at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2682) ~[hadoop-common-3.1.1.3.1.4.26-3.jar:?]

I think the issue here is that this error occurred too late - based on the stack trace, LLAP has already accepted/registered the fragment. The subsequent cleanup of this fragment/attempt also affects the first request. Which results in the LLAP crash described in HIVE-23061:

2020-03-17T06:49:11,304 ERROR [ExecutionCompletionThread #0 ()] org.apache.hadoop.hive.llap.daemon.impl.LlapDaemon: Thread Thread[ExecutionCompletionThread #0,5,main] threw an Exception. Shutting down now...
java.lang.IllegalStateException: Cannot invoke unregister on an entity which has not been registered
    at com.google.common.base.Preconditions.checkState(Preconditions.java:508) ~[hive-exec-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.32-1]
    at org.apache.hadoop.hive.llap.daemon.impl.QueryInfo$FinishableStateTracker.unregisterForUpdates(QueryInfo.java:256) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.QueryInfo.unregisterFinishableStateUpdate(QueryInfo.java:209) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.QueryFragmentInfo.unregisterForFinishableStateUpdates(QueryFragmentInfo.java:166) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService$TaskWrapper.maybeUnregisterForFinishedStateNotifications(TaskExecutorService.java:1177) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService$InternalCompletionListener.onSuccess(TaskExecutorService.java:980) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at org.apache.hadoop.hive.llap.daemon.impl.TaskExecutorService$InternalCompletionListener.onSuccess(TaskExecutorService.java:944) ~[hive-llap-server-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.26-3]
    at com.google.common.util.concurrent.Futures$CallbackListener.run(Futures.java:1021) ~[hive-exec-3.1.0.3.1.4.26-3.jar:3.1.0.3.1.4.32-1]
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) ~[?:1.8.0_191]
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) ~[?:1.8.0_191]
    at java.lang.Thread.run(Thread.java:748) [?:1.8.0_191]

Issue Links

• is duplicated by: HIVE-23061 LLAP crash due to unhandled exception: Cannot invoke unregister on an entity which has not been registered
• is related to: HIVE-23061 LLAP crash due to unhandled exception: Cannot invoke unregister on an entity which has not been registered