Skip to main content

CI/CD for Databases on GitLab

GitLab is a popular, open-source alternative to GitHub. In addition to a self-hosted version, GitLab also offers a hosted version at gitlab.com. Similar to GitHub, GitLab offers users storage for Git repositories, issue tracking, and CI/CD pipelines.

In this guide we will demonstrate how to use GitLab CI and Atlas to setup CI pipelines for your database schema changes.

Prerequisites​

Installing Atlas​

To download and install the latest release of the Atlas CLI, simply run the following in your terminal:

curl -sSf https://atlasgo.sh | sh

Installation instructions can be found here.

After installing Atlas locally, you will need to log in to your organization. You can do this by running the following command:

atlas login

Creating a bot token​

In order to report the results of your CI runs to Atlas Cloud, you will need to create a bot token for Atlas Cloud to use.

Follow these instructions to create a token and copy it.

Next, in your Gitlab project go to Settings -> CI/CD -> Variables and create a new variable called ATLAS_CLOUD_TOKEN. Paste your token in the value field.

Creating a variable for your database URL​

To avoid having plain-text database URLs which may contain sensitive information in your configuration files, create another variable named DB_URL and populate it with the URL (connection string) of your database.

To learn more about formatting URLs for different databases, see the URL documentation.

Creating a Gitlab access token (optional)​

Atlas will need permissions to comment lint reports on merge requests. To enable it, in your Gitlab project go to Settings -> Access Tokens. Create a new token. The role field should be set to "Reporter" or higher, and the "API" checkbox should be checked. Copy the token, and then go to Settings -> CI/CD -> Variables and create a new variable called GITLAB_TOKEN. Paste the token in the value field.

Choose a workflow​

Atlas supports two types of schema management workflows:

  • Versioned Migrations - In this flow, changes to the schema are defined as migrations (SQL scripts) and applied in order to reach the desired state.
  • Declarative Migrations - In this flow, the desired state of the database is defined as code and Atlas is responsible to calculate the migration plan to apply it.

To learn more about the differences and tradeoffs between these approaches, see the Declarative vs Versioned article.

Versioned Migrations Workflow​

In the versioned workflow, changes to the schema are represented by a migration directory in your codebase. Each file in this directory represents a transition to a new version of the schema.

Based on our blueprint for Modern CI/CD for Databases, our pipeline will:

  1. Lint new migration files whenever a merge request (MR) is opened.
  2. Push the migration directory to the Schema Registry when changes are merged to the mainline branch.
  3. Apply new migrations to our database.

Pushing a migration directory to Atlas Cloud​

Run the following command from the parent directory of your migration directory to create a "migration directory" repo in your Atlas Cloud organization (replace "app" with the name you want to give to your new repository):

$ atlas migrate push app \
  --dev-url "docker://postgres/15/dev?search_path=public"
info

If the migration directory contains multiple schemas, adjust the dev-url accordingly.

Atlas will print a URL leading to your migrations on Atlas Cloud. You can visit this URL to view your migrations.

Setting up GitLab CI​

Create a .gitlab-ci.yml file with the following pipelines, based on the type of your database. Remember to replace "app" with the real name of your repository.

.gitlab.yml
image: ubuntu:latest

services:
  - postgres:latest

variables:
  POSTGRES_DB: dev
  POSTGRES_USER: user
  POSTGRES_PASSWORD: pass

stages:
  - lint
  - push
  - apply

include:
  - component: $CI_SERVER_FQDN/arigaio/atlas/migrate-lint@~latest
    inputs:
      stage: lint
      dir: "file://migrations"
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN
      dir-name: "app"
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      gitlab-token: $GITLAB_TOKEN

  - component: $CI_SERVER_FQDN/arigaio/atlas/migrate-push@~latest
    inputs:
      stage: push
      branches:
        - main
      dir: "file://migrations"
      dir-name: "app"
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN
      
  - component: $CI_SERVER_FQDN/arigaio/atlas/migrate-apply@~latest
    inputs:
      stage: apply
      branches:
        - main
      dir: "file://migrations"
      url: $DB_URL
      revisions-schema: public
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN

Let's break down what this file is doing:

  1. The migrate-lint component will run on every new merge request. If new migrations are detected, Atlas will lint them and post the report as a merge request comment like this:
Screenshot of a GitLab merge request comment from Atlas, showing a successful linting report for a new database migration.
  1. After the merge request is merged into to main branch, the migrate-push component will push the new state of the schema to the Schema Registry on Atlas Cloud.
  2. Then, the migrate-apply component will deploy the new migrations to your database.

Testing our pipeline​

Let's take our pipeline for a spin:

  1. Locally, create a new branch and add a new migration with atlas migrate new --edit. Paste the following in th editor:
schema.sql
CREATE TABLE `test` (`c1` INT)
  1. Commit and push the changes.
  2. In Gitlab, open a merge request.
  3. View the lint report generated by Atlas. Follow the links to see the changes visually on Atlas Cloud.
  4. Merge the MR.
  5. When the pipeline is finished running, check your database to see if the changes were applied.

Declarative Migrations Workflow​

In the declarative workflow, developers provide the desired state of the database, as code. Atlas can read database schemas from various formats such as plain SQL, Atlas HCL, ORM models, and even another live database. Atlas then connects to the target database and calculates the diff between the current state and the desired state. It then generates a migration plan to bring the database to the desired state.

In this guide, we will use the SQL schema format.

Our goal​

When a merge request containing changes to the schema, we want Atlas to:

  • Compare the current state (your database) with the new desired state.
  • Create a migration plan show it to the user for approval.
  • Mark the plan as approved when the merge request is approved and merged.
  • During deployment, use the approved plan to apply the changes to the database.

Creating a simple SQL schema​

Create a file named schema.sql and fill it with the following content:

-- create table "users"
CREATE TABLE users(
	id int NOT NULL,
	name varchar(100) NULL,
	PRIMARY KEY(id)
);

-- create table "blog_posts"
CREATE TABLE blog_posts(
	id int NOT NULL,
	title varchar(100) NULL,
	body text NULL,
	author_id int NULL,
	PRIMARY KEY(id),
	CONSTRAINT author_fk FOREIGN KEY(author_id) REFERENCES users(id)
);

Then, create a configuration file for Atlas named atlas.hcl as follows:

atlas.hcl
env "gitlab" {
    url = getenv("DB_URL")
    schema {
        src = "file://schema.sql"
        repo {
            name = "app"
        }
    }
}

Pushing the schema to Atlas Cloud​

To push our initial schema to the Schema Registry on Atlas Cloud, run the following command:

$ atlas schema push app \
  --dev-url "docker://postgres/15/dev?search_path=public" \
  --env gitlab

Setting up GitLab CI​

Create a .gitlab-ci.yml file with the following pipelines, based on the type of your database.

.gitlab.yml
image: ubuntu:latest

services:
  - postgres:latest

variables:
  POSTGRES_DB: dev
  POSTGRES_USER: user
  POSTGRES_PASSWORD: pass

stages:
  - plan
  - push
  - apply

include:
  - component: $CI_SERVER_FQDN/arigaio/atlas/schema-plan@~latest
    inputs:
      stage: plan
      env: gitlab
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN
      gitlab-token: $GITLAB_TOKEN

  - component: $CI_SERVER_FQDN/arigaio/atlas/schema-push@~latest
    inputs:
      stage: push
      branches:
        - main
      env: gitlab
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      latest: true
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN

  - component: $CI_SERVER_FQDN/arigaio/atlas/schema-plan-approve@~latest
    inputs:
      stage: push
      branches:
        - main
      env: gitlab
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN

  - component: $CI_SERVER_FQDN/arigaio/atlas/schema-apply@~latest
    inputs:
      stage: apply
      branches:
        - main
      env: gitlab
      dev-url: "postgres://user:pass@postgres/dev?sslmode=disable"
      atlas-cloud-token: $ATLAS_CLOUD_TOKEN

  1. When a new merge request is opened, the schema-plan component will check if the desired state of the schema was changed. If it was, Atlas will generate a migration plan, lint it and post the report as a merge request comment.

  2. When the merge request is merged, two things happen: First, the updated schema is pushed to the schema registry by the schema-push component. Second, the plan that was created in the merege request will be approved.

  3. The schema-apply component will then be used to apply the new state of the schema to the database, using the plan that was just approved.

Testing our pipeline​

Let's see our CI/CD pipeline in action!

Step 1: make a schema change​

Let's add the "address" column to the users table:

schema.sql
-- create table "users"
CREATE TABLE users(
	id int NOT NULL,
	name varchar(100) NULL,
	address varchar(100) NULL,
	PRIMARY KEY(id)
);

-- create table "blog_posts"
CREATE TABLE blog_posts(
	id int NOT NULL,
	title varchar(100) NULL,
	body text NULL,
	author_id int NULL,
	PRIMARY KEY(id),
	CONSTRAINT author_fk FOREIGN KEY(author_id) REFERENCES users(id)
);

Now, let's commit the change to a new branch, push it to GitLab and open a merge request. The schema-plan component will use Atlas to create a migration plan from the current state of the database to the new desired state:

Screenshot of a GitLab merge request comment from Atlas, showing a successful linting report for a new database migration.

There are two things to note:

  • The comment also includes instructions to edit the plan. This is usefull when the plan has lint issues (for example, dropping a column will raise a "desctructive changes" error).
  • The plan is created in a "pending" state, which means Atlas can't use it yet against the real database.

Merging the changes​

Let's hit the merge button to merge the changes with the main branch. A new pipeline will be fired, with 3 jobs: The schema-plan-approve job will approve the plan that was generated earlier, the schema-push job will sync the new desired state in the schema registry, And then the schema-apply job will deploy the changes to our database.

Screenshot of a successful GitLab CI pipeline job, showing the 'schema-apply' step has completed.

The last thing to do is to inspect our database to make sure the changes were applied correctly:

$ atlas schema diff \
  --from $DB_URL \
  --to "file://schema.sql" \
  --dev-url "docker://postgres/15/dev?search_path=public"
Schemas are synced, no changes to be made.