Module 3 Lesson 4: Caching Dependencies

The #1 reason pipelines are slow is Downloading Libraries.

• Every build runs npm install or pip install.
• This takes 2 minutes every time. Cache allows you to "Save" these libraries between pipeline runs.

1. Cache vs. Artifacts

This is the most common confusion:

• Artifacts: Pass data forward to the next stage of the same pipeline. (e.g. Build -> Test).
• Cache: Pass data forward to the next pipeline (e.g. Today's run -> Tomorrow's run).

---

2. Setting Up the Cache

# Global cache for the whole project
cache:
  # Using the lock file as a key. 
  # If the lock file doesn't change, the cache remains valid!
  key:
    files:
      - package-lock.json
  paths:
    - node_modules/

test-job:
  script:
    - npm install # This will be INSTANT if the cache is hit
    - npm test

---

Visualizing the Process

graph TD
    Start[Input] --> Process[Processing]
    Process --> Decision{Check}
    Decision -->|Success| End[Complete]
    Decision -->|Retry| Process

---

3. Choosing the Cache Key

• Constant Key: key: project-wide. Every run uses the same cache. Simple, but can get "Dirty."
• Branch Key: key: $CI_COMMIT_REF_SLUG. Every branch has its own cache.
• Lock File Key (Pro): key: { files: [package-lock.json] }. The cache only updates when you actually add a new library.

---

4. Why Cache is "Fragile"

1. Not Guaranteed: Unlike Artifacts, GitLab might delete a cache if it needs space. Your script should always be able to run npm install from scratch if the cache is missing.
2. Storage Costs: Large caches (like 1GB of Docker layers) can fill up your runner's hard drive quickly.

---

Exercise: The Speed Benchmark

1. Create a project with a package.json that includes a heavy library (like lodash).
2. Run a pipeline that does npm install WITHOUT cache. Note the time.
3. Add the cache block from Section 2.
4. Run the pipeline again.
5. Run it a third time. Was the third run significantly faster than the first?
6. Modify the package-lock.json. Watch the logs: Does the runner say "Creating cache" or "Restoring cache"?

---

Summary

Caching is the difference between a pipeline that developers love (fast) and a pipeline that developers ignore (slow). By properly caching your vendor folders, you respect your developers' time and your company's compute budget.

Next Lesson: Logical control: Job control (only, except, rules).