Fix Docker configuration for Gitlab CI

[Andrew-Chen-Wang]

Description

• Just missing the services I believe according to @rocode?

This should provide the temporary fix for GitLab CI #2549

Rationale

Quick fix for those using GitLab CI

Use case(s) / visualization(s)

• I'm hoping to just completely eradicate the need for this setup with a follow-up PR for completely non-Docker approach on CI that supports celery, too.

[rocode]

Ah, note that you need the docker:dind service as well, or it will fail when trying to connect to the containers.

services:
    - docker:dind

[Andrew-Chen-Wang]

Thanks @rocode! Also, I've configured Travis to use a non Docker approach which cuts down the CI runtime by 1.5 minutes. Do you mind if I ping you again so that you can test that PR (when I find the time to copy and paste the config file and rewrite the tests) on Gitlab CI?

[rocode]

Thanks @rocode! Also, I've configured Travis to use a non Docker approach which cuts down the CI runtime by 1.5 minutes. Do you mind if I ping you again so that you can test that PR (when I find the time to copy and paste the config file and rewrite the tests) on Gitlab CI?

Sounds good! Our current CI is about ~5 minutes, so that will be very appreciated.

[Andrew-Chen-Wang]

@rocode Ah, I might as well get you started. My current Travis configuration is extremely wacky at the moment, but the gist of it, in Travis, is to use the built-in services. To avoid the entire Docker thing, you just need to add the environment variables that are in the .envs directory.

AFAIK in my config, that would be:

- USE_DOCKER="no"
- DATABASE_URL="postgres:///mydatabase"
- REDIS_URL="redis://localhost:6379/0"
- CELERY_BROKER_URL="redis://localhost:6379/0"

and my services are Postgres and Redis.

For celery, use this: - celery multi start beat worker -A config.celery_app --pool=solo

And your script should be pytest.