• config.yml vs gitlab-ci.yml
  • Jobs
  • Docker image definition
  • Workflows
    • Parallel and sequential job execution
    • Scheduled run
    • Manual run
  • Filter job by branch
  • Caching
• Contexts and variables
• Orbs
• Build environments
  • Machine and specific build environments

Migrating from CircleCI

If you are currently using CircleCI, you can migrate your CI/CD pipelines to GitLab CI/CD, and start making use of all its powerful features. Check out our comparison to see what’s different.

We have collected several resources that you may find useful before starting to migrate.

The Quick Start Guide is a good overview of how GitLab CI/CD works. You may also be interested in Auto DevOps which can be used to build, test, and deploy your applications with little to no configuration needed at all.

For advanced CI/CD teams, custom project templates can enable the reuse of pipeline configurations.

If you have questions that are not answered here, the config.yml vs gitlab-ci.yml

CircleCI’s config.yml configuration file defines scripts, jobs, and workflows (known as “stages” in GitLab). In GitLab, a similar approach is used with a .gitlab-ci.yml file in the root directory of your repository.

Jobs

In CircleCI, jobs are a collection of steps to perform a specific task. In GitLab, jobs are also a fundamental element in the configuration file. The checkout keyword is not necessary in GitLab CI/CD as the repository is automatically fetched.

CircleCI example job definition:

jobs:
  job1:
    steps:
      - checkout
      - run: "execute-script-for-job1"

Example of the same job definition in GitLab CI/CD:

job1:
  script: "execute-script-for-job1"

Docker image definition

CircleCI defines images at the job level, which is also supported by GitLab CI/CD. Additionally, GitLab CI/CD supports setting this globally to be used by all jobs that don’t have image defined.

CircleCI example image definition:

jobs:
  job1:
    docker:
      - image: ruby:2.6

Example of the same image definition in GitLab CI/CD:

job1:
  image: ruby:2.6

Workflows

CircleCI determines the run order for jobs with workflows. This is also used to determine concurrent, sequential, scheduled, or manual runs. The equivalent function in GitLab CI/CD is called stages. Jobs on the same stage run in parallel, and only run after previous stages complete. Execution of the next stage is skipped when a job fails by default, but this can be allowed to continue even after a failed job.

See the Pipeline Architecture Overview for guidance on different types of pipelines that you can use. Pipelines can be tailored to meet your needs, such as for a large complex project or a monorepo with independent defined components.

Parallel and sequential job execution

The following examples show how jobs can run in parallel, or sequentially:

1. job1 and job2 run in parallel (in the build stage for GitLab CI/CD).
2. job3 runs only after job1 and job2 complete successfully (in the test stage).
3. job4 runs only after job3 completes successfully (in the deploy stage).

CircleCI example with workflows:

version: 2
jobs:
  job1:
    steps:
      - checkout
      - run: make build dependencies
  job2:
    steps:
      - run: make build artifacts
  job3:
    steps:
      - run: make test
  job4:
    steps:
      - run: make deploy

workflows:
  version: 2
  jobs:
    - job1
    - job2
    - job3:
        requires:
          - job1
          - job2
    - job4:
        requires:
          - job3

Example of the same workflow as stages in GitLab CI/CD:

stages:
  - build
  - test
  - deploy

job1:
  stage: build
  script: make build dependencies

job2:
  stage: build
  script: make build artifacts

job3:
  stage: test
  script: make test

job4:
  stage: deploy
  script: make deploy

Scheduled run

GitLab CI/CD has an easy to use UI to schedule pipelines. Also, rules can be used to determine if jobs should be included or excluded from a scheduled pipeline.

CircleCI example of a scheduled workflow:

commit-workflow:
  jobs:
    - build
scheduled-workflow:
  triggers:
    - schedule:
        cron: "0 1 * * *"
        filters:
          branches:
            only: try-schedule-workflow
  jobs:
    - build

Example of the same scheduled pipeline using rules in GitLab CI/CD:

job1:
  script:
    - make build
  rules:
    - if: '$CI_PIPELINE_SOURCE == "schedule" && $CI_COMMIT_REF_NAME == "try-schedule-workflow"'

After the pipeline configuration is saved, you configure the cron schedule in the GitLab UI, and can enable or disable schedules in the UI as well.

Manual run

CircleCI example of a manual workflow:

release-branch-workflow:
  jobs:
    - build
    - testing:
        requires:
          - build
    - deploy:
        type: approval
        requires:
          - testing

Example of the same workflow using when: manual in GitLab CI/CD:

deploy_prod:
  stage: deploy
  script:
    - echo "Deploy to production server"
  when: manual

Filter job by branch

Rules are a mechanism to determine if the job runs for a specific branch.

CircleCI example of a job filtered by branch:

jobs:
  deploy:
    branches:
      only:
        - main
        - /rc-.*/

Example of the same workflow using rules in GitLab CI/CD:

deploy:
  stage: deploy
  script:
    - echo "Deploy job"
  rules:
    - if: '$CI_COMMIT_BRANCH == "main" || $CI_COMMIT_BRANCH =~ /^rc-/'

Caching

GitLab provides a caching mechanism to speed up build times for your jobs by reusing previously downloaded dependencies. It’s important to know the different between cache and artifacts to make the best use of these features.

CircleCI example of a job using a cache:

jobs:
  job1:
    steps:
      - restore_cache:
          key: source-v1-< .Revision >
      - checkout
      - run: npm install
      - save_cache:
          key: source-v1-< .Revision >
          paths:
            - "node_modules"

Example of the same pipeline using cache in GitLab CI/CD:

image: node:latest

# Cache modules in between jobs
cache:
  key: $CI_COMMIT_REF_SLUG
  paths:
    - .npm/

before_script:
  - npm ci --cache .npm --prefer-offline

test_async:
  script:
    - node ./specs/start.js ./specs/async.spec.js

Contexts and variables

CircleCI provides Orbs

There are two GitLab issues open addressing CircleCI Orbs and how GitLab can achieve similar functionality.

Build environments

CircleCI offers executors as the underlying technology to run a specific job. In GitLab, this is done by runners.

The following environments are supported:

Self-managed runners:

• Linux
• Windows
• macOS

GitLab.com shared runners:

• Linux
• Windows

Machine and specific build environments

Tags can be used to run jobs on different platforms, by telling GitLab which runners should run the jobs.

CircleCI example of a job running on a specific environment:

jobs:
  ubuntuJob:
    machine:
      image: ubuntu-1604:201903-01
    steps:
      - checkout
      - run: echo "Hello, $USER!"
  osxJob:
    macos:
      xcode: 11.3.0
    steps:
      - checkout
      - run: echo "Hello, $USER!"

Example of the same job using tags in GitLab CI/CD:

windows job:
  stage:
    - build
  tags:
    - windows
  script:
    - echo Hello, %USERNAME%!

osx job:
  stage:
    - build
  tags:
    - osx
  script:
    - echo "Hello, $USER!"