Troubleshooting segmentation fault (core dumped)

[ewag]

Hi,
I'm attempting to run a simple model forced by a large gridded reanalysis netcdf (~40 GB). When running Raven, the model gets as far as "Writing Output File Headers" and "Simulation Start..." before exiting with a "Segmentation Fault (core dumped)" error. Is this just because I don't have enough memory to run the model? Or is there some other approach I can take to troubleshoot this

I'm running Raven 3.5n on a Linux OS with 8gb of RAM.

Thanks!
-Erik

[ewag]

Followup: I have pared down my netcdf to a much more reasonable size and still get the error. Also tried compiling Raven from source on a different computer (MacOSX) with the same results.

[rchlumsk]

Hi Erik

Thanks for the question and update, trying a smaller netCDF file is a good step. Raven does have chunking options to read in the netCDF file in pieces, so with that feature the size of the netCDF file should be less of an issue overall.

I would suggest to try running Raven in :NoisyMode (add this to the rvi file), which might provide more of a clue to where it encounters an error. If you are comfortable with compiling Raven (which it sounds like you are), you can try adding some additional print statements with cout in the C++ code into a temporary Raven version, and use this version to try and see where the error is coming from.

My guess is that it is a problem with using the netCDF file somehow, but if you wanted to rule out other options, you could try replacing the netCDF forcing data with a single non-gridded forcing station temporarily to see if the issue is resolved. A few days of forcing data from a standard rvt gauge file would do.

Hope that helps and let us know if you are able to shed more light on this error. I am happy to keep corresponding and get this bug fixed.

Thanks,
Rob