Re: [Condor-users] output and input files

[Matthew Farrellee <matt@xxxxxxxxxx>]

On 08/18/2010 05:55 PM, Dave STREET wrote:
> I have a program the I want to run on condor for a user.
>
> but it has the following inbuilt file structure
>
> /executable
> /dependency1
> /dependency2
> /input/file1
> /input/file2
> /output/file1
> /output/file2
>
>
> my question has 2 parts
>
> first this is on a shared file system so I don't want to really have to
> transfer input files.
>
> can I use the following submit file? well should I be able to ?
>
> initialdir = /
> executable = executable
> transfer = no
>
> But my second and main question is how do I get the output files back?
>
> If I set the job to run once, I can using the local file system get it
> to out put to the /output/file1,2 whichs is cool
>
> but if I run the job twice it will over write the first output.
>
> What i really want is to use the condor submit file to somehow copy the
> output files back in to
>
> /output1/...
> /output2/...
> etc
>
> so use shared file system for the input files, and then set the
> initialdir = output$(process)/ and get condor to copy any changed files
> back to here.
>
> but it seems if I do this for the output it mucks up the input, and if
> set the input right it mucks up the output!!
>
> is there a way to do this ?
>
> Cheers
>
> Aaron

I would expect you to be able to specify the output files to transfer 
back, getting the benefit of input files in a shared filesystem and 
output files saved for multiple runs.

You might look at transfer_output_remaps.

Best,


matt