search

MAPC / carto-sql-api-heroku

Project for 1-click push deployment of the SQL API
BSD 3-Clause "New" or "Revised" License
0 stars 1 forks source link

Heroku deployment (12 factor app) #1

Open allthesignals opened 7 years ago

allthesignals commented 7 years ago

To deploy to Heroku, we need to make the config files pull from the environment.

https://12factor.net/config

Right now, they use templated files: https://github.com/MAPC/CartoDB-SQL-API/blob/master/config/environments/staging.js.example.

allthesignals commented 7 years ago

This comment suggests they're read from redis! -> https://github.com/CartoDB/CartoDB-SQL-API/blob/14846b254e4841c12afb14cf02a09bf9a4744216/config/environments/production.js.example#L23

allthesignals commented 7 years ago

Username is extracted from regex here đź‘Ť: https://github.com/CartoDB/CartoDB-SQL-API/blob/14846b254e4841c12afb14cf02a09bf9a4744216/config/environments/production.js.example#L17

allthesignals commented 7 years ago

Remaining mysteries, some I have emailed to @zingbot:

Which Redis database is the test suite pointing to? Documentation mentions this: “Make sure redis is running and knows about active cartodb user.” What is redis used for in the SQL API? What is an active cartodb user?

  • I made some progress on this front - the Redis store is used for configuration variables as well as setting up some of the user configuration (usernames). How should I create a “cartodb user” for development, staging, or production? I know the test suite does that with “cartodb250user”.
  • Related issue: https://github.com/codefornl/carto-sql-api-docker/issues/1 Would it be possible to remove the redis dependency? (In other words, do the core features of SQL API depend on redis?)

Carto official seems to use table schemas and associate them with specific Carto users. Do Carto usernames relate to the subdomain for SQL API specifically, and do table schemas in Postgres need to be tied to a specific CartoDB user? Our goal doesn’t include having any user management (only one main user to get it to work). For example: When I run node app.js test, I can visit this URL successfully: http://cartodb250user.localhost:8080/api/v2/sql?q=select%20*%20from%20test_geometry_table. I guess my question is how do I create or manage the “cartodb250user” part of the URL?

  • So far, I can see here that the regex looks for a username in Redis, then checks the Redis store for a hash called rails structured like this.
allthesignals commented 7 years ago

This shell script can be a way to generate the configuration file with server environment variables: https://github.com/codefornl/carto-sql-api-docker/blob/master/run.sh

allthesignals commented 7 years ago

Let's use postinstall in package.json to trigger the shell script before Heroku kicks off the Procfile: https://docs.npmjs.com/misc/scripts#examples.

allthesignals commented 7 years ago

Progress! Now to test on Heroku with .env variables configured correctly. One possible issue: if a user deploys without the proper environment variable, it's not clear if setting it in Heroku will kick off the package.json postinstall, which generates the configuration file. It might need a full rebuild.

allthesignals commented 7 years ago

^ On that front, according to docs:

prestart, start, poststart: Run by the npm start command.

allthesignals commented 7 years ago

Can't use those npm hooks because Procfile command needs environment variable