GitLab CI/CD Architecture Overview
What You'll Learn
You'll understand how GitLab's integrated CI/CD system works, including runners and how pipelines respond to repository events.
The Core Architecture
GitLab CI/CD is deeply integrated into the GitLab platform itself—unlike GitHub Actions, which feels like a separate feature, GitLab treats CI/CD as a first-class citizen of the repository experience.
Three Essential Components
GitLab Server (Control Plane)
The GitLab instance (self-hosted or GitLab.com) contains your repositories and acts as the orchestrator. When repository events occur (pushes, merge requests, tags), GitLab reads the .gitlab-ci.yml file at the repository root and schedules pipeline jobs.
Runners (Execution Engines) Runners are separate agents that pull jobs from GitLab and execute them. Think of runners as workers waiting at a job board—they check in with GitLab asking "got any work for me?" rather than GitLab pushing work to them. Runners can be:
- Shared runners (available to all projects)
- Group runners (scoped to project groups)
- Specific runners (dedicated to particular projects)
Pipeline Configuration (.gitlab-ci.yml)
This YAML file defines stages, jobs, and execution rules. GitLab parses this configuration when events trigger, creating a pipeline instance with scheduled jobs.
Event-Driven Execution
When you push code or create a merge request, GitLab immediately evaluates pipeline triggers defined in your configuration. Eligible jobs enter a queue, and available runners claim them based on tags and availability. This pull model allows elastic scaling—add more runners when workload increases.
Key Takeaway: GitLab CI/CD uses a pull-based architecture where runners actively claim jobs scheduled by the GitLab server in response to repository events, enabling flexible scaling and execution distribution.