Note: If you see a step below that could be improved (or is outdated), please update instructions. We rarely go through this process ourselves, so your fresh pair of eyes and your recent experience with it, makes you the best candidate to improve them for other users.
Install Postgres 9.x. Start the database server, if necessary.
For development, ensure that local connections do not require a password. Locate your pg_hba.conf file by
running ps aux | grep postgres and note the directory in the postgres or postmaster process, specified with -D.
It will look something like /Library/PostgreSQL/9.3/data. We'll call this the $POSTGRES_DATADIR. cd to $POSTGRES_DATADIR, and
edit pg_hba.conf to trust local socket connections and local IP connections. Restart postgres - on Mac OS X, there may be
restart scripts already in place with brew, if not use pg_ctl -D $POSTGRES_DATADIR restart.
Now, assuming the postgres database superuser is postgres, let's create the databases.
createdb -U postgres opencollective_localhost
createdb -U postgres opencollective_test
createuser -U postgres opencollective
psql -U postgres
> GRANT ALL PRIVILEGES ON DATABASE opencollective_localhost TO opencollective;
> GRANT ALL PRIVILEGES ON DATABASE opencollective_test TO opencollective;
- From the OpenCollective DropBox:
cp $DROPBOX/Engineering/config/DOTenv .env - There are other config files there, but for now they seem to be duplicated in
config
npm install
If you haven't already: export PATH=./node_modules/bin:$PATH. You probably want to add
that to your shell profile.
See Wiki.
All the calls to 3rd party services are stubbed using either sinon or nock.
If you get an error at the first test, you might have forgotten to run postgres. Use e.g. the following aliases to start/stop postgres:
export PGDATA='/usr/local/var/postgres'
alias pgstart='pg_ctl -l $PGDATA/server.log start'
alias pgstop='pg_ctl stop -m fast'
npm run start
Run the server on the side (in parallel because the reset scripts hits directly the api):
DEBUG=email npm run dev
Run the script afterwards:
npm run db:reset
You can now login on development (website repo) with [email protected] and the token will show up in your console.
You can auth to the paypal sandbox with [email protected] and paypal123.
Feel free to modify scripts/create_user_and_group.js to create your own user/group.
WIP
If you want to deploy to staging, you need to push your code to the staging branch. CircleCI will run the tests on this branch and push to Heroku for you if successful.
If you want to deploy the app on Heroku manually (only for production), you need to add the remotes:
git remote add heroku-production https://git.heroku.com/opencollective-prod-api.git
Then you can run:
git push heroku-production master
The tests delete all the opencollective_test database's tables and re-create them with the latest models.
For localhost or other environments, the migrations has to be run manually.
npm run migration:create
npm run db:migrate:dev
The migrations are run automatically for staging.
The migration script uses SEQUELIZE_ENV to know which Postgres config to take (check sequelize_cli.json). On staging and production, it will use PG_URL. We don't use NODE_ENV because heroku overrides the variable during the build process.
-
Push application with migration scripts to Heroku
-
heroku run bash -a opencollective-production-api -
npm run db:migrate