This is resolved. There was an issue in my build chain where spark 0.8.0 was getting built into the working application, and then spark 0.7.3 was getting built into the second (failing) application.
Frank Austin Nothaft [email protected] [email protected] 202-340-0466 On Dec 9, 2013, at 2:01 PM, Frank Austin Nothaft <[email protected]> wrote: > Hi all, > > I am getting a remote client shutdown error when running my application using > spark-0.8.0-incubating on EC2; Jey and I have been looking at this most of > today. My cluster is set up using the spark-ec2 scripts. We are running two > applications, one which binds successfully to the Spark master and runs. The > second application does not bind to the master, and throws the following > error: > > 2013-12-09 21:40:09 ERROR Client$ClientActor:64 - Connection to master > failed; stopping client > 2013-12-09 21:40:09 ERROR SparkDeploySchedulerBackend:64 - Disconnected from > Spark cluster! > 2013-12-09 21:40:09 ERROR ClusterScheduler:64 - Exiting due to error from > cluster scheduler: Disconnected from Spark cluster > > Even after this failure, I am still able to run the first command again. In > the application, we've put logging to debug and get no additional information > on the cause of the error. We also tried setting the Spark master logging > level to debug, but do not see any increase in logging. > > Through jdb, we traced this to > org.apache.spark.deploy.client.Client$ClientActor.receive, where the > RemoteClientShutdown case matches, and markDisconnected is called. However, > we have not been able to debug further. Any advice would be greatly > appreciated. > > Regards, > > Frank Austin Nothaft > [email protected] > [email protected] > 202-340-0466 > >
