A small disk fills faster than you think

A single pipeline run moved disk usage from 52.8% to 60.6% on a 6.6 GB root volume — roughly 500 MB per build.

Every run writes a fresh set of image layers, and nothing removes the old ones.

Why it is nasty

At that rate it is about five builds to a full disk. It does not announce itself as "out of space" either — it surfaces as confusing, unrelated-looking build failures.

The fix

df -h /
docker system df
docker image prune -f

I added the prune to the pipeline's post { always { … } } block so it runs after every build regardless of outcome.

Takeaway

Storage is a resource your pipeline consumes. Budget for it explicitly.