[Swift-devel] Re: swift + condor-g

[Mihael Hategan]

On Mon, 2009-04-13 at 20:01 +0000, Ben Clifford wrote:
> It has more serious problems (as in problems that stop it working for me 
> entirely) that are perhaps more important to address. The changes I made 
> to get it going with condor-g (but some are to make it work with 
> condor-anything) are in here:
> 
> http://www.ci.uchicago.edu/~benc/tmp/condor-g.patch
> 
> In summary what I changed to make it run at UJ are:
> 
>  * exit code handling - the in-SVN version expects an exit code file to 
> appear but never gets one.

Odd. It was supposed to use the exit code ad.

>  the patch removes that expectation - I haven't 
> looked enough to see if it returns exit codes properly or not because 
> Swift doesn't care.
> 
>  * LeaveJobInQueue needs to have no + at the start of it (I can't find 
> much clear documentation about this on the condor web site)
> 
>  * quoting is kinda screwy (as seems traditional with condor) - what 
> breaks running first.swift is for an empty argument, he preent quoting 
> behaviour makes the swift wrapper look forinput file called "" rather than 
> taking no input files.

What version of condor is that?

> 
> The patch above contains some other mess, as well as the changes I made to 
> get it to submit to the OSG CE at UJ.
> 
> 
> 
> On Mon, 13 Apr 2009, Mihael Hategan wrote:
> 
> > That provider I think needs some updating to use job logs instead of the
> > current hold-in-queue-and-then-remove scheme.
> > 
> > On Mon, 2009-04-13 at 19:36 +0000, Ben Clifford wrote:
> > > after more fiddling than I was expecting with the cog condor provider and 
> > > the condor installation that I'm using, I managed to get first.swift 
> > > running through condor-g onto the University of Johannesburg site that I 
> > > am working at.
> > > 
> > > so this looks hopeful.
> > > 
> > 
> >