when run nf in a culster, let's say gridengine, nf script dispatch jobs to multiple nodes via scheduler. Is it possible to use local storage for middle results (i.e. NXF_WORK), while Input/Output (raw data and result files) of the pipeline stays in share folders?
seems to me the work folder as well as all the sub-hash-folders are created at the beginning of the script on submit node only, not on job nodes. thus if using local storage, other job nodes will not be able to visit.