dbGaP Job Submission
All dbGaP data is stored on Isilon share mounted automatically when a user requests an interactive dbGaP session.
Step 1: Request an interactive session on Oscar on the dbGaP.
Group membership fordbgap, dbg_PiLastName, dbg_import, dbg_export
along with SLURM associations are required for working with dbGaP.
Step 2: Understand the dbGaP Data Hierarchy
The native GPFS file system (Home, Scratch, Data, Runtime) will be read-only access. All dbGaP related work must be done in the/dbGaP
root directory. The /dbGaP
directory has two sub-directories data
& results
. The downloaded datasets from the xfer server will be written to/dbgap/import/user
and all output files will be written to the/dbgap/results/username
directory. The hierarchy structure is:
Step 3: Submitting a dbGaP batch jobs
Home, Scratch & Data will be in read-only mode only users must write their code files, batch scripts in their designated. Example batch script for dbGaP jobs
Any output files including SLURM out and err files must be written to /dbGaP/results
directory.
Last updated