[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Submit files
- Date: Wed, 12 Dec 2007 18:00:45 -0000
- From: "Kewley, J \(John\)" <j.kewley@xxxxxxxx>
- Subject: Re: [Condor-users] Submit files
> If you can relax the requirements a little and only need to
> recover which OS the job was run under, then you can add the
> $$(OpSys) to one of the filenames being returned (log file,
> or otherwise). This will result in some empty garbage files
> with $$(OpSys) in the name on the submit node; however, when
> Condor runs the job and returns the new filenames will have
> the expanded OS string.
For the types of reports (governmental) that these guys were discussing,
"These jobs were run on a random mixture of LINUX and WINNT51" (and
reading between the lines,
"which it would be hard to reproduce exactly") is far less strong than
"These jobs were run on WINNT51 machines" (and we could reproduce the
run exactly by
providing the same random number seeds).
I think that is their point. Knowing on which OpSys and Arch each one
ran is great
for "Audit", but not really for "Reproducibility" and certainly not for
even if they could prove their codes were *identical* on different OSes.