On Tue, Dec 04, 2007 at 12:49:56PM +0200, Eugeny N Dzhurinsky wrote: > On Mon, Dec 03, 2007 at 09:38:57AM -0800, Owen O'Malley wrote: > > > > On Dec 3, 2007, at 8:59 AM, Eugeny N Dzhurinsky wrote: > > > >> Hello there! > >> > >> We would like to start several map/reduce jobs on the same host from > >> several > >> threads using different input and output dirs. However we are not able to > >> do > >> this for some reason, and we are getting the exception: > >> > >> 07/12/03 18:53:57 ERROR impl.OurTask: java.io.IOException: Target > >> /tmp/hadoop-bofh/mapred/local/localRunner/job_local_1.xml already exists > > The problem is that you are using the local runner. For multiple threaded > > access, the distributed mode is a better direction. (ie. Bring up a job > > tracker, some set of task trackers, and submit your job to the job > > tracker.) If you want to use local runner, you need to set mapred.local.dir > > differently on each process.) A fix for HADOOP-1733 would also fix this > > problem. > > I am using the code listed below to submit the job. I made mapred.local.dir to > point to different locations for different jobs (UUID is unique)
Nevermind, I just checked out the latest snapshot from trunk for hadoop 0.16, and looks like local jobs works just fine! Thank you! -- Eugene N Dzhurinsky
pgpPrkXWoMheP.pgp
Description: PGP signature
