[GE users] Loose vs. tight MPICH integration

Reuti reuti at staff.uni-marburg.de
Tue Apr 18 16:48:07 BST 2006


Hi,

Am 18.04.2006 um 17:39 schrieb Duong Ta:

> Davide,
>
> That's why I'm confused now: even if I use loose integration, MPI  
> processes are still under the control of SGE daemons (I'm using  
> MPICH). I'm not sure where did I do wrong :(

so you would prefer using Loose Integration on your system. Would you  
please post your loose PE definition and a ps output like you  
mentioned below. The SGE installation is shared across all nodes and  
not a local one? Which rsh command is compiled into your MPICH  
installation?

-- Reuti


> Best regards,
> Duong
>
> On 4/18/06, davide cittaro <daweonline at gmail.com> wrote: As written  
> in the howto page, using tight integration you have correct
> accounting and processes/directories controlled by SGE, definitely
> removed at the end of the job, while if you only enable loose
> integration, processes are started but they are under LAM daemons
> control only...
>
> d
>
> On 4/18/06, Duong Ta <duongtnb at gmail.com> wrote:
> > Dear,
> >
> > Anyone could tell me the differences in the procedures for loose  
> and tight
> > integration of MPICH to SGE v6? I've noticed there're a few  
> differences as
> > below:
> >
> > - Tight integration: we have the -catch_rsh for startmpi.sh. Loose
> > integration doesn't have.
> > - Tight integration: control slave = true, job is first task =  
> false. The
> > opposites are applied for loose integration
> >
> > However, when I check the output of "ps -e f -o  
> pid,ppid,pgrp,command", it
> > seems that there were no difference in running MPI jobs under  
> loose or tight
> > SGE integration. That means in both cases, I've always noticed  
> that the MPI
> > jobs were under the control of sge_shepherd. The resource usage  
> e.g., CPU
> > was also correctly reflected in both loose and tight integration.
> >
> > Any suggestions?
> >
> > Best regards,
> > Duong
> >
>
>
> --
> dawe
> http://dawe.ilbello.com
> ---
> "Prediction is very difficult, especially if it's about the future." -
> Niels Bohr
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list