Basic question about sharing docker instances

[Lucas Vazquez]

Hello everyone!

Today I discovered bitbucket pipelines and I find it very interesting.

But I have a basic performance problem, probably very common.

I have a script that looks like the following code:

image: docker/compose:1.29.2

pipelines:
  default:
    - step:
        name: Setup
        script:
          - cp .env.local .env
          - docker-compose up -d
          - echo Setup done
        services:
          - docker
    - step:
        name: Run migrations
        script:
          - docker-compose run project alembic upgrade head
          - echo Migrations done
        services:
          - docker
    - step:
        name: Run tests
        script:
          - docker-compose run project python -m pytest -rP -vv -x
          - echo Tests done
        services:
          - docker

What is the problem? The second step returns Error: No such container
I realized I need to setup the container for each step, but there is any way to do this?

This is a good way to implement what I want??

image: docker/compose:1.29.2

pipelines:
  default:
    - step:
        name: Run migrations
        script:
          - cp .env.local .env
          - docker-compose up -d
          - docker-compose run project alembic upgrade head
          - echo Migrations done
        services:
          - docker
    - step:
        name: Run tests
        script:
          - cp .env.local .env
          - docker-compose up -d
          - docker-compose run project python -m pytest -rP -vv -x
          - echo Tests done
        services:
          - docker

Just for context: the docker compose has a mysql, a firestore emulator, and an image that installs certain python packages.

[Aron Gombas _Midori_ · Accepted solution]

Bitbucket Pipelines has a feature called cache which allows one step to produce an artifact which can used by later steps:

https://support.atlassian.com/bitbucket-cloud/docs/cache-dependencies/

[Lucas Vazquez]

Thanks! This is what I was looking for

[Erez Maadani]

You should consider that each step runs on a different container, possibly on different host.

The simplest option would be to upload the created image (in step #1) to a repository and pull it / use it as a runner in the next step.