Disposable environments for tests.
WHAT THEY PROVIDE
A real database, message broker or cache for tests, created and destroyed per run.
WHY THAT BEATS SUBSTITUTES
Tests exercise the actual behaviour, including its quirks.
WHAT IT COSTS
Slower tests than in-memory substitutes.
WHAT TO USE WHERE
Substitutes for unit tests Real dependencies for integration tests
WHAT TO ENSURE
Each test run starts from a known state.
HOW
A fresh container, or a reset between runs.
WHAT TO BE CAREFUL WITH
Tests depending on each other's data Ports conflicting when runs overlap
WHAT TO DO ABOUT PORTS
Let the system assign one, and read it.
WHAT TO WAIT FOR
The dependency being ready, not merely started.
WHY
Tests otherwise fail intermittently, which is worse than failing consistently.
WHAT TO USE
A readiness check, with a timeout.
WHAT TO RUN IN THE PIPELINE
The same setup as locally.
WHY
Divergence produces failures that only appear in one place.
WHAT TO CLEAN UP
Containers and volumes, after every run.
WHY
Build agents otherwise fill their disks.
WHAT TO TEST AGAINST
The built image, not the source tree.
