Next: References
Up: Running on the
Previous: Running batch jobs
This section provides some basic troubleshooting information for
the model, and is still being written.
- Job won't start: No files.
If no new mpmm.error.* or mpmm.out.* files were been
created in the current working directory. Retry. If
failure recurs, check the following items. As a last resort, contact
system support staff, as it may be some problem with the system software
preventing the job from starting. Be sure to tell support as much about
the problem as you possibly can. If mpmm.log was created by the
script, you may want to send this to support as well.
- Executable file in place with executable permissions?
- Processor specification file (fort.50) in place?
- Does batch script change directories to the current working dir?
- Any messages in mpmm.log that may indicate source of problem?
- Job starts and dies: Files created. At least some of the
mpmm.out and mpmm.error files were created.
- Did the job run out of time?
- Did output file space fill up?
- Check the mpmm.out and mpmm.error files for an indication of
the problem.
- Check the mpmm.log file for an indication of a run-time crash
(such as a segmentation fault, bus error, or floating point exception).
- Job runs but CFL violation messages in one or more mpmm.out files.
- Possible causes;
- Problems with input data.
- Domain specifications do not match input data.
- Job runs but NaN's in model output.
- Possible causes;
- Problems with input data.
- Domain specifications do not match input data.
- Output from old runs. The mpmm.out and mpmm.error files
contain output from previous runs. Always delete these before each new
run.