Re: [Condor-users] Why parallel job won't run

["Kewley, J \(John\)" <j.kewley@xxxxxxxx>]

From: condor-users-bounces@xxxxxxxxxxx [mailto:condor-users-bounces@xxxxxxxxxxx]On Behalf Of Remon van Holstein

> I have a question i made a simple parallel submit job like this 
 
> universe = parallel
> executable = /bin/sleep
> arguments = 30
> machine_count = 4
 
> ( target.Arch == "INTEL" ) && ( target.OpSys == "LINUX" ) &&

I presume this is a REQUIREMENTS statement, but it seems incomplete.
 
> transfer_input_files = /bin/sleep

You shouldn't need to transfer /bin/sleep, there should be one on unix
machines already. Having said that, how does condor cope with a full pathname
like this, does it tray and *overwrite* /bin/sleep at the remote end?

There is also a transfer_executable which would appear more appropriate,
although it may behave a little differently to a full path, but I haven't 
tried it so don't know exactly.

> should_transfer_files = yes
> when_to_transfer_output = on_exit
> queue 
 
> But when I submit it it stays idle when I change the universe in vanilla the job runs but > when I put 100 behind queue then the job turns idle can somebody tell what the problem is 
 
Sorry, I can't see any other problems (since REQUIREMENTS is incomplete,
maybe there is something else missing). What do the log files say?
(note that with vanilla, you should probably remove machine_count)

Cheers

JK