Conda package update

[delgadom]

Workflow

• [x] Closes issues #80, #81, #82, #83, and #96
• [x] Includes PR #88
• [x] Passes travis tests
• [x] Image deployed on dev branch (see notebook and worker dev tags)
• [ ] Worker passes manual deployment test on compute.rhg

Compute-test.rhg.com functionality tests

• [ ] Notebook+worker passes test-hub deployment
• [ ] Notebook supports ipywidgets on test-hub
• [ ] Notebook:worker pairing tested with xarray+dask+holoviews distributed computing
• [ ] Notebook:worker pairing tested with dask.dataframe+parquet+datashader distributed computing
• [ ] Notebook:worker pairing tested with dask.dataframe+parquet new column distributed write
• [ ] dask_kubernetes graceful autoscale up & down
• [ ] extra_pip_packages and extra_conda_packages on worker
• [x] R kernel on notebook
• [x] Octave kernel on notebook
• [ ] conda:env:octave ipython kernel : octave-worker oct2py distributed octave run

Downstream images tested

• [ ] pyTC
• [ ] kerry_rainfall_model

Superuser signoff

• [ ] Kelly McCusker
• [ ] Ian Bolliger
• [ ] Ali Hamidi

(feel free to add yourself to this list)

Summary

Update a bunch of packages, including:

• pandas, numpy
• xarray, zarr, zict, netcdf4, netcdf-fortran
• dask, distributed, dask-searchcv, dask-ml
• holoviews, geoviews, pyviz, pyviz-comms

Incorporates new conda environments:

• R (moved from root env to new r env)
• octave

Adds apt-get packages:

• octave
• kubectl

Does not include updates to jupyterlab (see #85) - this introduced some bugs in the cluster interface. Additional exploration of different jupyterlab deployments is required.

Also creates a third image, octave-worker, with a minimal dask-worker build alongside an octave kernel. To use, pull rhodium/octave-worker:dev. This worker is meant to be paired with the octave environment's python kernel (rather than that environment's octave kernel), using dask.distributed to connect to the oct2py octave driver on the worker.

[delgadom]

The version of pandas in this build breaks the dev (vectorized) version of geopandas.

[delgadom]

Numpy int length conflict has been re-introduced. Likely due to a higher version of numpy relative to pandas.

[bolliger32]

superseeded by #102