hdrake / big-data-tutorial Goto Github PK

The datasets for the LLC4320 output that we use in the tutorial notebook have unlabelled coordinates i and j standing in for longitude and latitude, and the values of the coordinates are simply the indices. Meanwhile, we have access to a separate data set which contains all of the LLC4320 grid information.

It should be straight-forward to merge the two data sets and swap the XC / XG longitude coordinates for the i coordinate and the YC / YG latitude coordinates for the j coordinate, but I am not sure how to do this or if it is really this easy... One should look more closely at xmitgcm documentation (there may be an example out there that does precisely this) and the related gcmfaces package.

This is the paper reference for the llc4320 grid (which I generated) and explains how the face business works : https://doi.org/10.5194/gmd-8-3071-2015

Will send a PR adding it to the Notebook if that's ok

Currently, dask_kubernetes is not supported in our binder instances. I completely forgot to include it in the binder/environment.yml file and it is possible that simply adding the package will work right out of the box and there is nothing else to do here.

I suspect however (and remember reading this somewhere, but can't find it now) that installing dask_kubernetes will not work unless we specifying the proper configuration options for the particular Pangeo BinderHub resources. I already have included the binder/start and .dask/config.yaml files described in the Pangeo Binder documentation, but they don't do anything at the moment because I forgot the dask_kubernetes package.

E.g. see https://github.com/hdrake/big-data-tutorial/blob/master/notebooks/Lecture04_dask_and_big_data-serial.ipynb