I am running Shark/Spark (0.9.1) on Amazon EC2 using the supplied setup scripts. I am reading data out of S3 and then trying to write back a table into S3. The data can be read from S3 fine (so my credentials are correct) but when I try to write data to S3 I get the following error:
14/07/31 16:42:30 INFO scheduler.TaskSetManager: Loss was due to java.lang.IllegalArgumentException: Wrong FS: s3n://id:key@shadoop/tmp/hive-root/hive_2014-07-31_16-39-29_825_6436105804053790400/_tmp.-ext-10000, expected: hdfs://ecmachine.compute-1.amazonaws.com:9000 [duplicate 3]
I've tried several different methods of writing out data/tables but they all produce the same error. This particular error is generated from a HQL query like:
INSERT OVERWRITE DIRECTORY 's3n://id:key@shadoop/bucket' SELECT * FROM table;
Any ideas on why S3 is seen as a "wrong FS"?