RESCOMP Archives

March 2008

RESCOMP@LISTSERV.MIAMIOH.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"Woods, David M. Dr." <[log in to unmask]>
Reply To:
Research Computing Support <[log in to unmask]>, Woods, David M. Dr.
Date:
Tue, 25 Mar 2008 10:13:03 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (96 lines)
Pieter,

  Change "ppn=4" to "ppn=2" (you may also want to increase the value for "node").  Ppn stands for "processors per node" and all of our nodes are dual processor nodes.

Dave
________________________________________
From: Research Computing Support [[log in to unmask]] On Behalf Of Pieter Pelser [[log in to unmask]]
Sent: Tuesday, March 25, 2008 9:24 AM
To: [log in to unmask]
Subject: Re: Mb-Parallel Jobs

Dear Robin et al.,

Thank you for your advice. How do I change my batch script to do this?
If I do it like this, I get an error:

#!/bin/bash -l
#PBS -N mb-parallel
#PBS -l nodes=8:ppn=4
#PBS -q par_restr
#PBS -q batch
#PBS -m abe

module load mrbayes-3.1.2
mpiexec mb-mpi SelectedMB.nex > SelectedMB.out

This is the error message: "qsub: Job rejected by all possible
destinations"

I have a few jobs that will take a long time to run, but there is no
way of knowing how long. The analyses will run faster if I can use
more nodes, but that comes at the expense of having to wait longer
until these nodes are all available. Can you advise me on what the
best strategy would be?

Thanks,

Pieter

*******************************************************************
Dr. Pieter B. Pelser

Miami University - Botany Department
316 Pearson Hall
Oxford, Ohio 45056
USA

Email: [log in to unmask]
IChat: [log in to unmask]
Tel. office: +15135294270
Tel. cell: +15133303952
Fax: +15135294243
http://pieterpelser.blogspot.com/
*******************************************************************



On Mar 21, 2008, at 1:23 PM, Robin, Robin wrote:

> Pieter,
>
>    Dave is out these couple weeks.
>
>    You submitted 3 jobs. 2 of them went over the time were requested
> and
> thus stopped running. I increased the time requested for you for the
> 3rd job
> to be 480 hours.
>
>    If you expect your job to be finished that within 20 days, your job
> script should include "-l walltime=480:00:00"
>
>    Better still, you are also allowed to submit jobs that have
> *unlimited*
> time; if in the job script you submit it without the walltime option
> and use
> the "-q par_restr" in the job script.
>
>    I think when you are unsure of the time it could take to finish,
> it's
> safe to submit it with "-q par_restr".
>
>    Please let us know by sending to [log in to unmask] and we'll be
> happy
> to create the job script and modify the job script for you with the
> proper
> time that you expect it to run. Also, it will be helpful to let us
> know
> where your job script is at.
>
>    Your 3rd parallel MB job has been modified to run up to 480 hours.
>
> Thanks,
> Robin
>

ATOM RSS1 RSS2