Hi I'm running storm-yarn on CDH5.3.2 and it's working great, except that it exits seemingly at random after running for between a few hours and a few days.

I see an exception in the yarn log, followed by a graceful shutdown:

15/04/16 10:58:15 ERROR yarn.MasterServer: Unhandled error in AM:
org.apache.hadoop.security.token.SecretManager$InvalidToken: Invalid AMRMToken from appattempt_1429008207550_0003_000002
              at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
              at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57)
              at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
              at java.lang.reflect.Constructor.newInstance(Constructor.java:526)
              at org.apache.hadoop.yarn.ipc.RPCUtil.instantiateException(RPCUtil.java:53)
              at org.apache.hadoop.yarn.ipc.RPCUtil.unwrapAndThrowException(RPCUtil.java:104)
              at org.apache.hadoop.yarn.api.impl.pb.client.ApplicationMasterProtocolPBClientImpl.allocate(ApplicationMasterProtocolPBClientImpl.java:79)
              at sun.reflect.GeneratedMethodAccessor2.invoke(Unknown Source)
              at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
              at java.lang.reflect.Method.invoke(Method.java:606)
              at org.apache.hadoop.io.retry.RetryInvocationHandler.invokeMethod(RetryInvocationHandler.java:187)
              at org.apache.hadoop.io.retry.RetryInvocationHandler.invoke(RetryInvocationHandler.java:102)
              at com.sun.proxy.$Proxy12.allocate(Unknown Source)
              at org.apache.hadoop.yarn.client.api.impl.AMRMClientImpl.allocate(AMRMClientImpl.java:278)
              at com.yahoo.storm.yarn.MasterServer$1.run(MasterServer.java:69)
Caused by: org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.security.token.SecretManager$InvalidToken): Invalid AMRMToken from appattempt_1429008207550_0003_000002
              at org.apache.hadoop.ipc.Client.call(Client.java:1411)
              at org.apache.hadoop.ipc.Client.call(Client.java:1364)
              at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:206)
              at com.sun.proxy.$Proxy11.allocate(Unknown Source)
              at org.apache.hadoop.yarn.api.impl.pb.client.ApplicationMasterProtocolPBClientImpl.allocate(ApplicationMasterProtocolPBClientImpl.java:77)
              ... 8 more
15/04/16 10:58:15 INFO yarn.StormMasterServerHandler: stopping supervisors...
15/04/16 10:58:15 INFO yarn.StormMasterServerHandler: stopping UI...
15/04/16 10:58:15 INFO yarn.StormMasterServerHandler: stopping nimbus...

This will be a blocking issue for deploying Cloudera. Other options include Hortonworks who claim to support storm-yarn but we have significant CDH knowledge and don't want to switch, or stand-alone storm which we do have experience with but would rather run a single cluster.

Any help would be awesome,

Thanks,

James