Hi Connell It looks like a bug in hive, I checked with hive 0.9 . If you are loading data from local fs to hive tables using 'LOAD DATA LOCAL INPATH' and if a file with the same name exists in the table's location then the new file will be suffixed by *_copy_1.
But if we do the 'LOAD DATA INPATH' for a file in hdfs then there is no rename happening but just a move task is getting triggered. Since a file with same name exists in same hdfs location, hadoop fs move operation throws an error. I'll file a jira for this issue and update the same here. Regards Bejoy KS ________________________________ From: "Connell, Chuck" <[email protected]> To: "[email protected]" <[email protected]> Sent: Wednesday, July 25, 2012 8:51 PM Subject: Problem replacing existing Hive file with modified copy I created a Hive table that consists of two files, names1.txt and names2.txt. The table works correctly and answers all queries etc. I want to REPLACE names2.txt with a modified version. I copied the new version of names2.txt to the /tmp/input folder within HDFS. Then I tried the command: hive -e "load data inpath '/tmp/input/names2.txt' into table names" I get an error: "Failed with exception null. Exception error, return code 1 from org.apache.hadoop.hive.ql.exec.MoveTask" Is this just a bug in this version (hadoop-hive-0.7.1+42.43-2.noarch, via CDH3 distro)? Is my syntax correct to update a file in Hive? TIA, Chuck Connell Nuance R&D Data Team Burlington, MA
