srothmei
commented
9 years ago Hi John,
OAR is indeed a point which causes headache, especially because the documentation is not very good.
However, i assume you have already tried to simply replace each occurrence of "oarsub" within the shell scripts by your job sheduler? What was the issue with this? I haven't tried this myself yet but from a first thought this should work because no OAR-specific aspects where used in the pipeline. Job-dpendencies for example are realized through the file system itself i.e. by checking the current number of stored result-files etc.
To run the pipeline on a (local) desktop computer, one can try to remove the oar call around the actual command encapsulated in tick marks. E.g. oarsub -n TestJob -l walltime=12:00:00 "./doMyTracking.sh" -> ./doMyTracking.sh
Best, Simon
Hi.
The pipeline is excellent and I'd very much like to use it. However both myself and our IT administrator have found OAR a headache to install. It's also not straightforward how to run the pipeline locally on a single machine.
I can see that it should be possible to run most of the code without using OAR. However at same time it looks like most of it is quite tightly mixed up with the oarsub commands.
Would it be possible for you to provide a set of stand-alone bash or whatever scripts that simply list the analysis steps and function calls sequentially?
I should then be able to adapt this according to the more conventional job submission systems and parallelization systems we already have available locally. Also being able to do local test-runs without cluster access would be very useful for further development and testing.
I hope this isn't a big ask.
Thanks,
j