Gitlab pipeline: How to recache node modules only when dependency changed? - continuous-integration

Question

I am working on the performance tuning for Gitlab pipeline using cache.
This is a nodejs project using npm for the dependency management. I have put the node_modules folder into cache for subsequent stages with following setting:
build:
stage: build
only:
- develop
script:
- npm install
cache:
key: $CI_COMMIT_REF_SLUG
paths:
- node_modules/
Could I make the cache available for pipeline triggered next time? Or the cache is accessible in single pipeline?
If I can access that within multiple pipeline, could I recache the node module only when we change package.json?

Answer 1

First, put the cache on the global level. This will make sure, that the jobs share the same cache.
Second, you can use cache:key:files introduced with GitLab 12.5 to only recreate the cache when the package.json changes.
cache:
key:
files:
- package.json
paths:
- node_modules/
build:
stage: build
only:
- develop
script:
- npm install
Further information:
https://docs.gitlab.com/ee/ci/yaml/#cachekeyfiles
Additional hints:
You might want to check on package-lock.json instead of package.json.
I recommend reading the cache mismatch chapter in the documentation to make sure you don't run into common problems where the cache might not be restored.
Instead of simply adding npm install, you can also skip this step when the node_modules folder was recreated from cache. Following bash addition to your npm install will only run the command, if the node_modules folder doesn't exist.
build:
stage: build
only:
- develop
script:
- if [ ! -d "node_modules" ]; then npm install; fi