Closed carlestena closed 5 years ago
You have to build the netcdf-c and hdf5 libraries with MPI enabled (see https://www.unidata.ucar.edu/software/netcdf/docs/getting_and_building_netcdf.html#build_parallel).
Thanks @jswhit. I know how to install HDF5 and NEtCDF4-C in bash, but not using conda. Is to make easier the installation of my future package
I don't either. There may already be conda packages though (https://anaconda.org/spectraldns/libnetcdf-parallel and https://anaconda.org/spectralDNS/hdf5-parallel)
I've already tried to use hdf5-parallel and libnetcdf-parallel.
But I cannot tested it because I have another dependency: esmpy[version='>=7.1.0r'] -> esmf==7.1.0r -> hdf5[version='>=1.10.2,<1.10.3.0a0']
and the hdf5-parallel is <= 1.8.17
As a temporary work around, using conda, I downgraded netcdf4 to version 1.2.4. which seems to be linked to parallel libraries. As a side-effect, this also downgrade numpy. You can still check if it still match your dependencies.
But i thought that parallel netcdf starts at 1.3.1 version
hope you got this straightened out, closing now...
I'm having that error while trying to run my job when installing netcdf4 using conda: ValueError: parallel mode requires MPI enabled netcdf-c
my yml file is using that (and more much libraries) channels:
dependencies:
So, Have I to add --enable-parallel in somewhere? Will I have the same problem with HDF5?