Configuring Databases

This document describes how to use the official CircleCI pre-built Docker container images for a database service in CircleCI 2.0.

PostgreSQL database testing example
Optional customization

Overview

CircleCI provides pre-built images for languages and services like databases with a lot of conveniences added into the images on CircleCI Docker Hub.

The following example shows a 2.0 .circleci/config.yml file with one job called build. Docker is selected for the executor and the first image is the primary container where all execution occurs. This example has a second image and this will be used as the service image. The first image is the programming language Python. The Python image has pip installed and -browsers for browser testing. The secondary image gives access to things like databases.

PostgreSQL database testing example

In the primary image the config defines an environment variable with the environment key, giving it a URL. The URL tells it that it is a PostgreSQL database, so it will default to the PostgreSQL default port. This pre-built circleci image includes a database and a user already. The username is postgres and database is circle_test. So, you can begin with using that user and database without having to set it up yourself.

Set the POSTGRES_USER environment variable in your CircleCI config to postgres to add the role to the image as follows:

      - image: circleci/postgres:9.6-alpine
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
        environment:
          POSTGRES_USER: postgres

This Postgres image in the example is slightly modified already with -ram at the end. It runs in-memory so it does not hit the disk and that will significantly improve the testing performance on this PostgreSQL database by using this image.

version: 2
jobs:
  build:

    # Primary container image where all commands run

    docker:
      - image: circleci/python:3.6.2-stretch-browsers
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
        environment:
          TEST_DATABASE_URL: postgresql://root@localhost/circle_test

    # Service container image

      - image: circleci/postgres:9.6.5-alpine-ram
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference

    steps:
      - checkout
      - run: sudo apt-get update
      - run: sudo apt-get install postgresql-client-9.6
      - run: whoami
      - run: |
          psql \
          -d $TEST_DATABASE_URL \
          -c "CREATE TABLE test (name char(25));"
      - run: |
          psql \
          -d $TEST_DATABASE_URL \
          -c "INSERT INTO test VALUES ('John'), ('Joanna'), ('Jennifer');"
      - run: |
          psql \
          -d $TEST_DATABASE_URL \
          -c "SELECT * from test"

The steps run checkout first, then install the Postgres client tools. The postgres:9.6.5-alpine-ram image doesn’t install any client-specific database adapters. For example, for Python, you might install psycopg2 so that you can interface with the PostgreSQL database. See Pre-Built CircleCI Services Images for the list of images and for a video of this build configuration.

In this example, the config installs the PostgreSQL client tools to get access to psql. Note: that sudo is run because the images do not run under the root account like most containers do by default. CircleCI has a circle account that runs commands by default, so if you want to do admin privileges or root privileges, you need to add sudo in front of your commands.

Three commands follow the postgresql-client-9.6 installation that interact with the database service. These are SQL commands that create a table called test, insert a value into that table, and select from the table. After committing changes and pushing them to GitHub, the build is automatically triggered on CircleCI and spins up the primary container.

Note: CircleCI injects a number of convenience environment variables into the primary container that you can use in conditionals throughout the rest of your build. For example, CIRCLE_NODE_INDEX and CIRCLE_NODE_TOTAL are related to concurrent build environments. See the Build Specific Environment Variables document for details.

When the database service spins up, it automatically creates the database circlecitest and the root role that you can use to log in and run your tests. It isn’t running as root, it is using the circle account. Then the database tests run to create a table, insert value into the table, and when SELECT is run on the table, the value comes out.

Optional customization

This section describes additional optional configuration for further customizing your build and avoiding race conditions.

Optimizing Postgres images

The default circleci/postgres Docker image uses regular persistent storage on disk. Using tmpfs may make tests run faster and may use fewer resources. To use a variant leveraging tmpfs storage, just append -ram to the circleci/postgres tag (i.e., circleci/postgres:9.6-alpine-ram).

PostGIS is also available and can be combined with the previous example: circleci/postgres:9.6-alpine-postgis-ram

Using binaries

To use pg_dump, pg_restore and similar utilities requires some extra configuration to ensure that pg_dump invocations will also use the correct version. Add the following to your config.yml file to enable pg_* or equivalent database utilities:

     steps:
    # Add the Postgres 9.6 binaries to the path.
       - run: echo 'export PATH=/usr/lib/postgresql/9.6/bin/:$PATH' >> $BASH_ENV

Using Dockerize to wait for dependencies

Using multiple Docker containers for your jobs may cause race conditions if the service in a container does not start before the job tries to use it. For example, your PostgreSQL container might be running, but might not be ready to accept connections. Work around this problem by using dockerize to wait for dependencies. Following is an example of how to do this in your CircleCI config.yml file:

version: 2.0
jobs:
  build:
    working_directory: /your/workdir
    docker:
      - image: your/image_for_primary_container
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
      - image: postgres:9.6.2-alpine
        auth:
          username: mydockerhub-user
          password: $DOCKERHUB_PASSWORD  # context / project UI env-var reference
        environment:
          POSTGRES_USER: your_postgres_user
          POSTGRES_DB: your_postgres_test
    steps:
      - checkout
      - run:
          name: install dockerize
          command: wget https://github.com/jwilder/dockerize/releases/download/$DOCKERIZE_VERSION/dockerize-linux-amd64-$DOCKERIZE_VERSION.tar.gz && sudo tar -C /usr/local/bin -xzvf dockerize-linux-amd64-$DOCKERIZE_VERSION.tar.gz && rm dockerize-linux-amd64-$DOCKERIZE_VERSION.tar.gz
          environment:
            DOCKERIZE_VERSION: v0.3.0
      - run:
          name: Wait for db
          command: dockerize -wait tcp://localhost:5432 -timeout 1m

It is possible to apply the same principle for the following databases:

MySQL:

dockerize -wait tcp://localhost:3306 -timeout 1m

Redis:

dockerize -wait tcp://localhost:6379 -timeout 1m

Redis also has a CLI available:

sudo apt-get install redis-tools ; while ! redis-cli ping 2>/dev/null ; do sleep 1 ; done

Other services such as web servers:

dockerize -wait http://localhost:80 -timeout 1m