Just a theory... But it could be that nohup starts the command in a separate shell. If that is the case then your env variables might change. i.e. HADOOP_HOME so may be the nohup version is using an altogether different hadoop installation (perhaps just a local install).
Havent tested above.. but just an idea. -Vinay ________________________________ From: Ashish Thusoo <[email protected]> To: "[email protected]" <[email protected]> Sent: Friday, August 21, 2009 3:51:07 PM Subject: RE: Nohupping the hive CLI causes wrong behavior Sorry. Did not see this. Please disregard my previous mail. Ashish -----Original Message----- From: Steven Wong [mailto:[email protected]] Sent: Thursday, August 20, 2009 7:20 PM To: [email protected] Subject: RE: Nohupping the hive CLI causes wrong behavior The M/R job that's generated for the nohup case is very different from the non-nohup case: It has only 1 map task and it ran to completion very quickly. The M/R job for non-nohup case has many map tasks and it took a while to run (to generate the correct result). -----Original Message----- From: Raghu Murthy [mailto:[email protected]] Sent: Thursday, August 20, 2009 15:42 To: [email protected] Subject: Re: Nohupping the hive CLI causes wrong behavior Did you check that nohup.out didn't have the result? On 8/20/09 3:21 PM, "Steven Wong" <[email protected]> wrote: > I get correct result if I run: hive -e "select count(1) from blah blah" > > But I get no result if I run: nohup hive -e "select count(1) from blah blah" > > In both cases, hive exited with 0 (success). Very strange. > > Hive is from trunk revision 795559. > > Steven
