YCozy created YARN-10232:
----------------------------
Summary: InvalidStateTransitionException: Invalid event:
LAUNCH_FAILED at RUNNING
Key: YARN-10232
URL: https://issues.apache.org/jira/browse/YARN-10232
Project: Hadoop YARN
Issue Type: Bug
Reporter: YCozy
We were testing YARN under network partition and found the following ERROR in
RM's log.
{code:java}
2020-04-11 13:10:39,739 ERROR
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl:
App attempt: appattempt_6_0001_000002 can't handle this event at current state
org.apache.hadoop.yarn.state.InvalidStateTransitionException: Invalid event:
LAUNCH_FAILED at RUNNING
at
org.apache.hadoop.yarn.state.StateMachineFactory.doTransition(StateMachineFactory.java:305)
at
org.apache.hadoop.yarn.state.StateMachineFactory.access$500(StateMachineFactory.java:46)
at
org.apache.hadoop.yarn.state.StateMachineFactory$InternalStateMachine.doTransition(StateMachineFactory.java:487)
at
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl.handle(RMAppAttemptImpl.java:916)
at
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl.handle(RMAppAttemptImpl.java:121)
at
org.apache.hadoop.yarn.server.resourcemanager.ResourceManager$ApplicationAttemptEventDispatcher.handle(ResourceManager.java:1097)
at
org.apache.hadoop.yarn.server.resourcemanager.ResourceManager$ApplicationAttemptEventDispatcher.handle(ResourceManager.java:1078)
at
org.apache.hadoop.yarn.event.AsyncDispatcher.dispatch(AsyncDispatcher.java:222)
at
org.apache.hadoop.yarn.event.AsyncDispatcher$1.run(AsyncDispatcher.java:138)
at java.lang.Thread.run(Thread.java:748)
{code}
After analyzing the logs, we have recovered the triggering process of this bug:
* We have a cluster with one RM and one NM.
* A client tries to start a YARN service.
* RM send a request to NM to start the containers:
NM's log:
{code:java}
2020-04-11 14:23:44,030 INFO SecurityLogger.org.apache.hadoop.ipc.Server: Auth
successful for appattempt_6_0001_000002 (auth:SIMPLE)
2020-04-11 14:23:44,229 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl:
Start request for container_6_0001_02_000001 by user appattempt_6_0001_000002
{code}
* NM starts the containers successfully:
NM's log:
{code:java}
2020-04-11 14:23:44,347 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.application.ApplicationImpl:
Application application_6_0001 transitioned from INITING to RUNNING
2020-04-11 14:23:44,357 INFO
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerImpl:
Container container_6_0001_02_000001 transitioned from NEW to LOCALIZING
{code}
* However, due to network partition, NM failed to send back the RPC response.
* After a while, the application is running happily:
RM's log:
{code:java}
2020-04-11 14:23:50,359 INFO
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl:
appattempt_6_0001_000002 State change from ALLOCATED to RUNNING on event =
REGISTERED
020-04-11 14:23:50,359 INFO
org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl:
application_6_0001 State change from ACCEPTED to RUNNING on event =
ATTEMPT_REGISTERED{code}
* Then, since RM didn't receive the RPC response for startContainers, it
retries:
NM's log:
{code:java}
2020-04-11 14:23:54,392 INFO SecurityLogger.org.apache.hadoop.ipc.Server: Auth
successful for appattempt_6_0001_000002 (auth:SIMPLE)
{code}
* But since the attempt is actually running, this launch request does not
succeed:
NM's log:
{code:java}
2020-04-11 14:23:54,401 ERROR
org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl:
Unauthorized request to start container.
Attempt to relaunch the same container with id container_6_0001_02_000001.
{code}
RM's log:
{code:java}
2020-04-11 14:23:54,428 INFO
org.apache.hadoop.yarn.server.resourcemanager.amlauncher.AMLauncher: Error
launching appattempt_6_0001_000002. Got exception:
org.apache.hadoop.yarn.exceptions.YarnException: Unauthorized request to start
container.
Attempt to relaunch the same container with id container_6_0001_02_000001.
at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
at
sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
at
sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
at
org.apache.hadoop.yarn.api.records.impl.pb.SerializedExceptionPBImpl.instantiateExceptionImpl(SerializedExceptionPBImpl.java:171)
at
org.apache.hadoop.yarn.api.records.impl.pb.SerializedExceptionPBImpl.instantiateException(SerializedExceptionPBImpl.java:182)
at
org.apache.hadoop.yarn.api.records.impl.pb.SerializedExceptionPBImpl.deSerialize(SerializedExceptionPBImpl.java:106)
at
org.apache.hadoop.yarn.server.resourcemanager.amlauncher.AMLauncher.launch(AMLauncher.java:137)
at
org.apache.hadoop.yarn.server.resourcemanager.amlauncher.AMLauncher.run(AMLauncher.java:344)
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748){code}
* As a result, RM gets the following InvalidStateTransitionException ERROR:
RM's log:
{code:java}
2020-04-11 14:23:54,429 ERROR
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl:
App attempt: appattempt_6_0001_000002 can't handle this event at current state
org.apache.hadoop.yarn.state.InvalidStateTransitionException: Invalid event:
LAUNCH_FAILED at RUNNING
at
org.apache.hadoop.yarn.state.StateMachineFactory.doTransition(StateMachineFactory.java:305)
at
org.apache.hadoop.yarn.state.StateMachineFactory.access$500(StateMachineFactory.java:46)
at
org.apache.hadoop.yarn.state.StateMachineFactory$InternalStateMachine.doTransition(StateMachineFactory.java:487)
at
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl.handle(RMAppAttemptImpl.java:916)
at
org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl.handle(RMAppAttemptImpl.java:121)
at
org.apache.hadoop.yarn.server.resourcemanager.ResourceManager$ApplicationAttemptEventDispatcher.handle(ResourceManager.java:1097)
at
org.apache.hadoop.yarn.server.resourcemanager.ResourceManager$ApplicationAttemptEventDispatcher.handle(ResourceManager.java:1078)
at
org.apache.hadoop.yarn.event.AsyncDispatcher.dispatch(AsyncDispatcher.java:222)
at
org.apache.hadoop.yarn.event.AsyncDispatcher$1.run(AsyncDispatcher.java:138)
at java.lang.Thread.run(Thread.java:748)
{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
