Executing in unique directories

You can instruct Condor to read input files from and write output files to more than one directory.

To direct Condor to use individual directories for reading and writing files in an iterative process, first create the directories. Use a consistent naming convention for each directory, and include in the names the full range of process numbers that you plan to execute.

For example, to execute a program four times and use individual directories for each of the four executions, create four directories. Use a naming convention that includes the full range of process IDs for four executions, 0 - 3. You can use any naming convention that you choose. In this example, you might name your directories dir0 - dir3.

The submit file for this example contains attributes and commands that instruct the Condor system to perform one executable four times. The following attributes direct Condor to perform each of the program executions within individual directories:

InitialDir = dir0 # This directory is used for job <cluster ID>.0
Queue
InitialDir = dir1 # This directory is used for job <cluster ID>.1
Queue
InitialDir = dir2 # This directory is used for job <cluster ID>.2
Queue
InitialDir = dir3 # This directory is used for job <cluster ID>.3
Queue

A shorter way to do the same thing is to use the following:

InitialDir = dir$(Process)
Queue 4

You then can place an input file for individual executions within each directory.