Hi Reuti, You are completely right, I realised that I had a mistake in my configuration so it wasn't working properly.
Thank you very much again. On Wed, Apr 30, 2014 at 10:14 PM, Reuti <[email protected]> wrote: > Am 30.04.2014 um 20:02 schrieb Yago Fernández Pinilla: > > > That´s exactly what I´m doing, but instead of using qrsh, I want to use > qsub, should it work either way? > > They are both in the game. > > First you submit your job like usual using `qsub`. On the master node of > the parallel job you should observe that the `mpiexec` is a kid of your > configured starter_mthod which in turn is a kid if sge_shepherd. > > On the slave nodes you should see something similar, but here the > processes are bound to your starter method, the `qrsh_starter` and the > sge_shepherd. > > As said: what did you set up in detail and what do you observe in your > cluster? > > -- Reuti > > > Thanks again > > > > Yago > > > > > > On Wed, Apr 30, 2014 at 7:53 PM, Reuti <[email protected]> > wrote: > > Am 30.04.2014 um 19:46 schrieb Yago Fernández Pinilla: > > > > > Hi again! > > > > > > Thanks a lot for replying. > > > > > > I think I already try this and it was just executed in the main node > but using a integration that mas configured by me for Platform MPI, but i > haven´t try with another MPI. Have you tried? > > > > I used Open MPI, but it's independent from the used MPI implementation > AFAICS. Also on the slave nodes the qrsh's kid is the defined starter > method. > > > > Do you observe, that the Platform MPI processes are bound to the > sge_shepherd (`pe -e f`)? > > > > You set "export MPI_REMSH=rsh" in your job script and defined a PE with > -catch_rsh to route the `rsh` to `qrsh -inherit ...`? > > > > -- Reuti > > > > > > > I checked the documentation but I couldn´t find anything about this, > did you read this somewhere? > > > > > > Thank you very much > > > > > > Yago > > > > > > > > > On Wed, Apr 30, 2014 at 12:56 PM, Reuti <[email protected]> > wrote: > > > Hi, > > > > > > Am 29.04.2014 um 21:09 schrieb Yago Fernández Pinilla: > > > > > > > Is it possible to execute a script in Grid Engine as a startup > script in every of the nodes with tight integration? > > > > > > > > The same as starter_method but instead of executing in the main > node in all the nodes. > > > > > > This is the default for a tight integration to use it for slave tasks > too. Do you observe something different? > > > > > > -- Reuti > > > > > > > > > > > > > > Thanks in advance > > > > > > > > Yago > > > > > > > > -- > > > > Yago Fernández Pinilla > > > > e-mail: [email protected] > > > > > > > > _______________________________________________ > > > > users mailing list > > > > [email protected] > > > > https://gridengine.org/mailman/listinfo/users > > > > > > > > > > > > > > > -- > > > Yago Fernández Pinilla > > > e-mail: [email protected] > > > > > > > > > > > > > -- > > Yago Fernández Pinilla > > e-mail: [email protected] > > > > -- Yago Fernández Pinilla e-mail: [email protected]
_______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
