[Dune-devel] [Dune-Commit] [Commit] dune-grid - c966253: [Tests] Added parallel grid tests using MPI if available

Dominic Kempf dominic.r.kempf at gmail.com
Tue Nov 3 21:20:07 CET 2015


Okay, Good to know. I guess the rest of the mess will be cleaned up by the
docker setup that we plan to use on our buildserver.

On Tuesday, November 3, 2015, Christian Engwer <
christian.engwer at uni-muenster.de> wrote:

> On Tue, Nov 03, 2015 at 08:42:18PM +0100, Dominic Kempf wrote:
> > Hey Christian,
> >
> > The killed process on timeout is the mpirun process, which should then
> kill
> > all child processes, or am I mistaken? In my quick test this morning it
> > worked.
>
> As far as I remember my own experiences, it only get a SIG15, but if
> the process hangs properly, you need a SIG9 to clean up.
>
> But as I said, I think this will happen very seldom.
>
> Ciao
> Christian
>
> > Best,
> > Dominic
> >
> > On Tuesday, November 3, 2015, Christian Engwer <
> > christian.engwer at uni-muenster.de <javascript:;>> wrote:
> >
> > > Hi Dominic,
> > >
> > > > You are of course right about the timeouts, it was an oversight by
> > > myself.
> > > > Fortunately,
> > > > the solution is easy. CMake tests have a TIMEOUT property, that takes
> > > > precedence over
> > > > the timeout settings of the ctest run (which have no timeout if run
> > > without
> > > > special options).
> > >
> > > unfortunately this might not be enough. Killing the rank-0 process is
> > > not necessarily enough to kill all mpi processes, depending on the
> > > type of bug you ran into.
> > >
> > > But of course the TIMEOUT option is better than nothing and should
> > > suffice in most cases.
> > >
> > > Ciao
> > > Christian
> > >
>
> --
> Prof. Dr. Christian Engwer
> Institut für Numerische und Angewandte Mathematik
> Fachbereich Mathematik und Informatik der Universität Münster
> Einsteinstrasse 62
> 48149 Münster
>
> E-Mail  christian.engwer at uni-muenster.de <javascript:;>
> Telefon +49 251 83-35067
> FAX     +49 251 83-32729
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dune-project.org/pipermail/dune-devel/attachments/20151103/6f443e57/attachment.htm>


More information about the Dune-devel mailing list