[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] custom directory for output files

> Hi Kewley,

You may call me John or JK :)
> Sorry for the unclear question.
> It was more like the latter, but more specialized:
> initialdir = <init-path>
> outputdir = <out-path>
> errordir = <err-path>
> and so on ....

I don't know a simple way to do that, but you can simulate some of that as follows:

You can use $(PROCESS) to match which job within a batch you are talking about.
Then, I think you can say:

# Ensure separate directory for error and output files
output = <out-path>/file.$(PROCESS)
error = <err-path>/file.$(PROCESS)

initialdir = <init-path>

Ramifications of this:
* <out-path> and <err-path> will need to be full paths so that initialdir isn't prepended
* Your standard input file will then have to live in <init-path> or specify
  its/their locations like error/output above, or another full path
* Any other input files (TRANSFER_INPUT_FILES) will need to be in <init-path>
* NOTE I had expected the executable to have to be there too, but a simple test 
  suggests it MUST be in the submitting directory (unless prefixed with an apropriate
  directory of course)
* Log file will go to <init-path> unless you redirect it as output and error above
* All output files will be transferred back to <init-path>

Hopefully that is a bit closer to what you want.

Note: above test done for vanilla job on non-NFS Linux machine, but basics should hold.