[torquedev] [torqueusers] Change in interactive job behavior from previous torque versions.

David Beer dbeer at adaptivecomputing.com
Thu Apr 15 12:01:46 MDT 2010


You're right. We'll try to be more open. Thanks for your help.

David

----- Original Message -----
> On Thu, Apr 15, 2010 at 12:56 PM, Justin Bronder < jsbronder at gmail.com
> > wrote:
> 
> 
> 
> 
> 
> On 14/04/10 13:19 -0400, Glen Beane wrote:
> > On Fri, Mar 12, 2010 at 6:45 PM, David Beer <
> > dbeer at adaptivecomputing.com >wrote:
> >
> > > I talked to the developers who worked on this, and they assumed
> > > that submitting qsub -I pbsi.sh would just ignore pbsi.sh, so they
> > > thought this
> > > wouldn't changed the behavior. I will try to resolve this.
> > >
> >
> >
> > this change in behavior is still catching torque users off guard. I
> > think the change should be backed out and the devs that added this
> > need to find a
> > different way to accomplish what they need (maybe a new command line
> > argument and not hijack the -I argument)
> >
> > -glen
> 
> For what it's worth, changes like this and other problems in the 2.4
> series were why I didn't even package it for Gentoo until 2.4.6. While
> I did
> finally release the package for testing, I will not push this on our
> stable users until there is some sense of coherence between Adaptive
> and the
> experienced users on this list.
> 
> It seems to me that if this is a feature (?) desired by some of the
> developers at Adaptive, then Glen has suggested the best solution. Are
> there any barriers to implementing it?
> 
> 
> 
> What is particularly disturbing is this was a documented (and used)
> feature of torque, and apparently it was assumed to be a bug. If the
> developers had consulted with the community they would have known that
> they were making a change that would impact a lot of long time
> TORQUE/PBS users. It should have been caught before making it into a
> production release.
> 
> The TORQUE project will continue to be plagued by these unannounced
> behavior changes/bugs until the development process is more
> transparent and open. The community NEEDS to be involved, or at some
> point someone is going to get sick of this and fork TORQUE.
> 
> _______________________________________________ torquedev mailing list
> torquedev at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torquedev

-- 
David Beer | Senior Software Engineer
Adaptive Computing



More information about the torquedev mailing list