Dockerfile for plotting maps in pace

[elynnwu]

Purpose

In order to utilize fv3viz on daint for plotting cubed-sphere tile data as maps with cartopy, we use a docker image to install all the requirements. Daint can run this through sarus.

Infrastructure changes:

• postprocessing.Dockerfile: this is the dockerfile used to make this image

Instead of installing fv3viz as a package, we install all the requirements and import the plotting function directly. To import cube plotting inside the container:

import sys
sys.path.append("/fv3net/external/fv3viz/")
from fv3viz import pcolormesh_cube

To plot on daint:

sarus pull elynnwu/pace:ubuntu18
srun -C gpu --partition=debug --account=# sarus run --mount=type=bind,source=/path/to/data/files,destination=/work elynnwu/pace:ubuntu18 python /work/example.py

Checklist

Before submitting this PR, please make sure:

• You have followed the coding standards guidelines established at Code Review Checklist.
• All relevant documentation has been updated or added (e.g. README, CONTRIBUTING docs)

[mcgibbon]

I'd strongly recommend pinning the commit of fv3net, the repo gives no stability guarantees.

If this is relevant now or in the future, we push images of fv3net to GCR every commit. I believe the image name is us.gcr.io/vcm-ml/fv3net.

[elynnwu]

launch jenkins

[FlorianDeconinck]

LGTM!
Don't hesitate to put more comments on those to describe the why of the packages, this is often a good idea for future reference when requirements evolves