Am 20.08.2013 um 20:03 schrieb Mark Hahn <h...@mcmaster.ca>: >> 1. Many people's job scripts use ssh either directly (to say clean up /tmp) >> or indirectly from mpirun. > > sure.
Well, GridEngine could catch the call to rsh/ssh (unless it uses an absolute path) and route it to GridEngine's `qrsh --inherit ...`, so they run under SGE's control and accounting. Also the creation and removal of the local scratch directory is done automatically. In case that more than one scratch directory is necessary for the job, a prolog/epilog can provide and remove it. Especially: if one deletes a job, you need proper trapping in the job script as otherwise the custom coded removal of the scratch direcory can't take place any longer - the job script might have been killed already. This could fill up the local disk over time. >> (good mpirun's use the batch engine's per-node >> daemon to launch the binaries not ssh). > > why should a scheduler have daemons cluttering up compute nodes? I think he refers to a daemon like sge_execd in SGE to receive jobs from the qmaster, and there will be similar ones for other queuing systems running on each node I would expect. -- Reuti > also, do you really launch so many very big but very short jobs > that startup time is a serious concern? I'm questioning assumptions here. > > thanks, mark. > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
