GitLab CI pipelines for a whole company

Their legacy CI system was a costly, daily puzzle. So they rewrote it with GitLab CI, cutting maintenance down to just eight hours a month.

#1about 3 minutes

Overcoming the challenges of a legacy Jenkins pipeline

The initial Jenkins setup suffered from an undocumented and unversioned Groovy codebase, leading to high maintenance toil for the operations team.

#2about 4 minutes

Establishing new requirements for a modern CI system

The decision to migrate was driven by high support costs, while retaining successful concepts like monorepos and setting new goals for simplicity and clarity.

#3about 3 minutes

How GitLab CI pipelines use stages, jobs, and DAGs

GitLab CI pipelines are structured with sequential stages and parallel jobs, which can be optimized into a Directed Acyclic Graph (DAG) using the `needs` keyword.

#4about 4 minutes

Applying principles for scalable and maintainable pipelines

Key principles for success include recycling configurations with `include`, documenting complex jobs, assigning clear ownership, and tracking known issues as pitfalls.

#5about 4 minutes

Customizing pipeline runtimes and implementing robust testing

Pipeline execution is customized using variable Docker images, while testing is handled with Docker Compose and dynamic environments are managed for quality assurance.

#6about 3 minutes

Solving performance bottlenecks from data sharing and dependencies

Sharing large artifacts like `node_modules` is slow, so performance is improved by implementing local package caches and runner-level volume mounts.

#7about 3 minutes

Key lessons learned from the CI migration process

Important takeaways include educating developers on best practices, automating dependency updates with Renovate, and carefully managing changes in a shared monorepo.

#8about 4 minutes

A feature wishlist for improving GitLab CI workflows

Future improvements to GitLab CI could include native HashiCorp Vault integration, map support for variables, simpler rule evaluation, and faster artifact sharing.