Yes, current releases do not have a way of prohibiting user-override of MCA params, so a user could indeed circumvent the directive to require an allocation. The original intent of the parameter was to close a hole that allowed users to mistakenly overload the head node of a cluster by forgetting to get an allocation prior to running the job. It was not intended to stop someone from deliberately violating that rule.
The devel trunk currently does contain the ability for a system admin to specify parameters that cannot be overridden by the user. This helps the situation, but a user could still build there own version of OMPI and bypass those specifications. Bottom line: if a user is determined to violate the rule, there isn't much we (the OMPI community) can do about it :-( On Tue, Jun 4, 2013 at 11:48 AM, Reuti <re...@staff.uni-marburg.de> wrote: > Am 04.06.2013 um 20:38 schrieb Ralph Castain: > > > There is an Mca param to require an allocation > > But this can be requested (or not) at execution time? > > Even a decicated compilation with a builtin test of an allocation won't > give the intended effect, as someone could use his own compilation of Open > MPI in his home directory with his own options - even the binaries could > have been compiled on a different machine (hence putting a test there > inside also won't work). > > Maybe a better way would be to check the running processes and remove > `mpiexec`s by a cronjob - unless some renames a copy of `mpiexec`... > > -- Reuti > > > > Sent from my iPhone > > > > On Jun 4, 2013, at 11:18 AM, Orion Poplawski <or...@cora.nwra.com> > wrote: > > > >> I'd like to be able to force mpirun to require being run under a > gridengine environment. Any ideas on how to achieve this, if possible? > >> > >> -- > >> Orion Poplawski > >> Technical Manager 303-415-9701 x222 > >> NWRA, Boulder/CoRA Office FAX: 303-415-9702 > >> 3380 Mitchell Lane or...@nwra.com > >> Boulder, CO 80301 http://www.nwra.com > >> _______________________________________________ > >> users mailing list > >> us...@open-mpi.org > >> http://www.open-mpi.org/mailman/listinfo.cgi/users > > > > _______________________________________________ > > users mailing list > > us...@open-mpi.org > > http://www.open-mpi.org/mailman/listinfo.cgi/users > > > _______________________________________________ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users >
