I had another question about the LOAD DATA LOCAL INPATH it does a exec.CopyTask first and then does a exec.MoveTask but I think it is leaving a file handle open because when I run a thread which bulk inserts data into partitions the open file count for the user under which my hive thrift server runs keeps on increasing till it reaches the max after which i start getting "connection refused".
I am currently still on hive 0.5.0 and have checked out the svn repository but can't exactly figure out the location where the files are being left open. anyone aware of similar problems ? Thanks Viral On Wed, Dec 15, 2010 at 3:51 PM, Leo Alekseyev <[email protected]> wrote: > You can use CREATE EXTERNAL TABLE... LOCATION. > > See http://wiki.apache.org/hadoop/Hive/LanguageManual/DDL and examples > thereof. When you LOAD DATA INPATH, the directory gets moved to the > Hive warehouse dir; it does not get modified. > > On Wed, Dec 15, 2010 at 3:33 PM, Mark <[email protected]> wrote: > > Can someone explain what actually happens when you create a table and > import > > data into using "LOAD DATA INPATH..." > > > > I noticed that when I load the data from files already existing in HDFS > that > > it actually removes the original file from its location and moves it > under > > the /user/hive directory. Is there anyway I can prevent this from > happening > > or is this just the way things work? At this point is the file modified > in > > anyway? I have some other Hadoop jobs that rely on this data. Should I > just > > update those jobs to operate on the data within these directories? Thanks > > >
