Containerize centillion

[charlesreid1]

Making centillion serverless (and solving #20, run centillion on heroku) is going to be much easier if we containerize centillion.

We've had excellent expereinces with jfloff's alpine-python-3, a super lightweight Python container that's easy to configure with a requirements.txt file. This is a good bet.

Should we split centillion into sub-services? My initial response is no, there's no good reason since everything operates as a single unit. But that's not quite true - the Flask application and the Search object are a single unit, but when re-indexing, the application spawns independent tasks - so there might be some usefulness to having a container to run the search engine, and another container to update the search index, with both containers bind-mounting the search_index folder.

Furthermore, if we end up using SqlAlchemy (is this even necessary if we're using a container?), this may be wise to run as a separate service in a separate container. (Then again, maybe not.)

Okay, let's summarize.

First steps to dockerize centillion:

• build a Dockerfile for whoosh/centillion and mayke shure it wurk guud
• build a single-service docker-compose.yml file (bind-mount site_index directory) to run as a pod
• test the configuration locally
• test the configuration using beta site
• if everything's going well, pull the old switcheroo with search.nihdatacommons.us

[charlesreid1]

Revisiting this.

We want to containerize centillion, but the motivation is for testing - we want to make testing easier for the version 1.7 release.

Test plan: https://github.com/dcppc/centillion/blob/testing/tests/Readme.md

QE plan: https://github.com/dcppc/centillion/blob/testing/quality/Readme.md

[charlesreid1]

Now that we have packaged centillion and released version 1.7, we are ready to Dockerize centillion. Let's pick this back up.