The difference between using the starter method and the shepherd wrapper I was bringing up is that the starter method is run by the shepherd and the shepherd already assumes a few things to be there, such as the job’s working directory. If any of those preconditions is not met then starting the job will fail and the starter method may not even get run. So if you want to do things in your pseudo-prolog that create a proper execution environment for the job in the first place then wrapping the shepherd may be the way to go.
Otherwise the starter method is probably the better approach as it is less prone to be destroyed when upgrading the Grid Engine version (for instance). Cheers, Fritz > Am 08.06.2016 um 13:56 schrieb Bill Bryce <bbr...@univa.com>: > > I could be wrong Chris but I think the prolog runs in a separate parent/child > process tree than the job. You could put this kind of functionality into a > starter_method where it does the NFS mounts and chroot actions then just runs > the job script. the $JOB_ID will be in the environment and you can just take > the rest of the command line arguments with $@ and pass them through. For an > example of ‘something similar’ not exactly what you asked about but similar, > have a look at Daniel Gruber’s blog entry: > http://gridengine.eu/index.php/grid-engine-internals/226-using-open-containers-with-runc-in-a-univa-grid-engine-compute-cluster-2014-06-28 > > I see that Fritz replied as well with a slightly different answer. > > Regards, > > Bill. > > > >> On Jun 8, 2016, at 7:00 AM, Chris Dagdigian <d...@sonsorol.org> wrote: >> >> >> Hey folks -- need my brain refreshed on prolog behavior ... >> >> Trying to figure out if a prolog script would be suitable for dramatically >> changing the execution environment -- doing things like NFS filesystem >> unmounts or chroot actions so that an incoming job would execute in the >> changed environment. >> >> I can see the prolog running as 'me' and as a child of the sge_shepherd >> daemon but I don't have enough of a test lab setup to confirm that the >> prolog is running on the execution host and if the parent/child process >> relationship is such that chroot jail actions performed by a prolog would be >> where the jobscript ends up running.... >> >> Anyone have a quick answer? Thanks! >> >> Chris >> >> >> >> _______________________________________________ >> users mailing list >> users@gridengine.org >> https://gridengine.org/mailman/listinfo/users > > William Bryce | VP Products > Univa Corporation, Toronto > E: bbr...@univa.com | D: 647-9742841 | Toll-Free (800) 370-5320 > W: Univa.com | FB: facebook.com/univa.corporation | T: twitter.com/Grid_Engine > > _______________________________________________ > users mailing list > users@gridengine.org > https://gridengine.org/mailman/listinfo/users Fritz Ferstl | CTO and Business Development, EMEA Univa Corporation <http://www.univa.com/> | The Data Center Optimization Company E-Mail: ffer...@univa.com | Phone: +49.941.5695.9351 | Mobile: +49.170.819.7390
_______________________________________________ users mailing list users@gridengine.org https://gridengine.org/mailman/listinfo/users
