Hello, I have a hadoop cluster upgraded to Hadoop 2.x and everything with it works fine. (Runs M/R jobs, able to perform actions on HDFS).
When i run a pig script using pig grunt shell or pig -x mapreduce -f 'test.pig'. In either case it connects to Hadoop cluster, starts M/R job, the M/R job completes fine. However the shell hangs and never returns. 2014-03-13 00:17:04,341 [JobControl] INFO org.apache.hadoop.mapreduce.Job - The url to track the job: https://apollo-jt.vip.org.com:50030/proxy/application_1394582929977_7433/ 2014-03-13 00:17:04,342 [main] INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher - HadoopJobId: job_1394582929977_7433 2014-03-13 00:17:04,342 [main] INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher - Processing aliases A,B,C 2014-03-13 00:17:04,342 [main] INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher - detailed locations: M: A[4,4],C[6,4],B[5,4] C: C[6,4],B[5,4] R: C[6,4] 2014-03-13 00:17:04,365 [main] INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher - 0% complete 2014-03-13 00:17:36,232 [main] INFO org.apache.hadoop.ipc.Client - Retrying connect to server: <AM_Host_Name>/<AM_Host_IP>:47718. Already tried 0 time(s); maxRetries=45 2014-03-13 00:17:36,232 [main] INFO org.apache.hadoop.ipc.Client - Retrying connect to server: <AM_Host_Name>/<AM_Host_IP>:47718. Already tried 1 time(s); maxRetries=45 2014-03-13 00:17:36,232 [main] INFO org.apache.hadoop.ipc.Client - Retrying connect to server: <AM_Host_Name>/<AM_Host_IP>:47718. Already tried 2 time(s); maxRetries=45 .. ... On analyzing, i found that AM_Host_Name matches the Application Master of the M/R job. Question 1) Does the client machine attempts to connect to Application Master, in order to get the status of M/R Job ? 2) If #1 is true, and since its Hadoop 2.x secure cluster, does it mean it requires firewall to be open between client and application master (any node in the cluster) and PORT ? 3) I assumed #1 and #2 are true and hence got the firewall to be opened between client and all nodes in hadoop cluster (since anyone can be application master) for port 47718. However to my surprise i found that this 47718 port changed. Is there a setting or a group of port numbers that are used to communicate between client and AM in order to report status ? If yes where can i find this list ? 4) How do i get the grunt shell back and see the status/progress of job from client machine ? -- Deepak
