One image, many settings.
WHAT THE PRINCIPLE IS
The same image runs everywhere, configured differently.
WHY
Anything else means the tested artefact is not the deployed one.
WHAT VARIES
Addresses of dependencies Credentials Resource sizes Replica counts Feature settings
WHAT TO AVOID
Copying manifests per environment.
WHY
They diverge, and nobody knows how.
WHAT TO USE INSTEAD
A base definition with per-environment overlays, or templating with values files.
WHAT AN OVERLAY APPROACH PROVIDES
Explicit differences, visible at a glance.
WHAT A TEMPLATING APPROACH PROVIDES
Parameterisation, and packaging for distribution.
WHAT EITHER REQUIRES
Everything in version control.
WHAT TO REVIEW
The rendered output, before applying.
WHY
It is what actually reaches the cluster.
WHAT TO KEEP IDENTICAL BETWEEN ENVIRONMENTS
The image The structure of the configuration Security settings
WHY SECURITY SETTINGS
Relaxing them in testing hides the problems that appear in production.
WHAT TO DIFFER DELIBERATELY
Scale, and external dependencies.
WHAT TO NEVER PUT IN THE REPOSITORY
Production secrets in plain text.
WHAT TO DOCUMENT
Which values differ, and why.
