[ 
https://issues.apache.org/jira/browse/TEZ-2855?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Siddharth Seth updated TEZ-2855:
--------------------------------
             Assignee: Siddharth Seth
    Affects Version/s:     (was: 0.8.0-alpha)
                       0.5.0
     Target Version/s: 0.7.1, 0.6.3, 0.8.1  (was: 0.8.1)

This goes all the way back to 0.5.
If a Vertex initialization is delayed - likely due to a large number of 
upstream vertices, and a task from a started vertex finishes very fast which 
generates an event for the uninitialized vertex - we try handling the event 
before the VM is setup.
InputInitializerEvents are not affected - since these events are cached while a 
vertex is in state NEW.

This was hit running LLAP unit tests - were task assignment and execution can 
be faster. The faster assignment and execution allows for the condition to be 
hit.
It is possible to hit this in regular jobs as well - but less likely since 
there's generally a delay in a container getting work. Hitting it in local mode 
is possible though. Targeting the fix up to 0.6.

> NPE while routing events
> ------------------------
>
>                 Key: TEZ-2855
>                 URL: https://issues.apache.org/jira/browse/TEZ-2855
>             Project: Apache Tez
>          Issue Type: Bug
>    Affects Versions: 0.5.0
>            Reporter: Siddharth Seth
>            Assignee: Siddharth Seth
>            Priority: Critical
>         Attachments: 2855log.gz
>
>
> Observed while running against 0.8.0-alpha. This will likely affect 0.7 as 
> well - that'll be known after debugging.
> {code}
> 2015-09-24T12:13:42,675 ERROR [Dispatcher thread: Central] 
> common.AsyncDispatcher: Error in dispatcher thread
> java.lang.NullPointerException
>   at 
> org.apache.tez.dag.app.dag.impl.VertexImpl.handleRoutedTezEvents(VertexImpl.java:4429)
>  ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.tez.dag.app.dag.impl.VertexImpl.access$4000(VertexImpl.java:203) 
> ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.tez.dag.app.dag.impl.VertexImpl$RouteEventTransition.transition(VertexImpl.java:4175)
>  ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.tez.dag.app.dag.impl.VertexImpl$RouteEventTransition.transition(VertexImpl.java:4167)
>  ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.hadoop.yarn.state.StateMachineFactory$MultipleInternalArc.doTransition(StateMachineFactory.java:385)
>  ~[hadoop-yarn-common-2.6.0.jar:?]
>   at 
> org.apache.hadoop.yarn.state.StateMachineFactory.doTransition(StateMachineFactory.java:302)
>  ~[hadoop-yarn-common-2.6.0.jar:?]
>   at 
> org.apache.hadoop.yarn.state.StateMachineFactory.access$300(StateMachineFactory.java:46)
>  ~[hadoop-yarn-common-2.6.0.jar:?]
>   at 
> org.apache.hadoop.yarn.state.StateMachineFactory$InternalStateMachine.doTransition(StateMachineFactory.java:448)
>  ~[hadoop-yarn-common-2.6.0.jar:?]
>   at 
> org.apache.tez.state.StateMachineTez.doTransition(StateMachineTez.java:57) 
> ~[TezAppJar.jar:0.8.0-alpha]
>   at org.apache.tez.dag.app.dag.impl.VertexImpl.handle(VertexImpl.java:1906) 
> ~[TezAppJar.jar:0.8.0-alpha]
>   at org.apache.tez.dag.app.dag.impl.VertexImpl.handle(VertexImpl.java:202) 
> ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.tez.dag.app.DAGAppMaster$VertexEventDispatcher.handle(DAGAppMaster.java:2069)
>  ~[TezAppJar.jar:0.8.0-alpha]
>   at 
> org.apache.tez.dag.app.DAGAppMaster$VertexEventDispatcher.handle(DAGAppMaster.java:2055)
>  ~[TezAppJar.jar:0.8.0-alpha]
>   at org.apache.tez.common.AsyncDispatcher.dispatch(AsyncDispatcher.java:183) 
> [tez-common-0.8.0-alpha.jar:0.8.0-alpha]
>   at org.apache.tez.common.AsyncDispatcher$1.run(AsyncDispatcher.java:114) 
> [tez-common-0.8.0-alpha.jar:0.8.0-alpha]
>   at java.lang.Thread.run(Thread.java:745) [?:1.8.0_40]
> 2015-09-24T12:13:42,681 INFO [HistoryEventHandlingThread] 
> impl.SimpleHistoryLoggingService: Writing event TASK_ATTEMPT_FINISHED to 
> history file
> {code}
> Looks like the VertexManager was null.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to