A container tag is a name that can point to different image content over time. A digest identifies specific content. Pinning a digest makes the selected base image explicit, but it also means a rebuilt application will keep using that image until the reference is changed.

Reproducible selection and timely patching are separate requirements. A deployment process needs to satisfy both by recording what it uses and providing a reviewed path for replacing it.

Know which image the build selected

Docker's build guidance explains that tags are mutable and that digest pinning fixes the image content used by a build. It also discusses the need to update pinned references when adopting newer base images. Docker's build practices describe this tradeoff.

A web service may need a base operating-system patch even when its application source has not changed. Rebuilding with the old pinned digest keeps the old package. Check the replacement image’s contents and digest as part of the patch.

Follow the patch through each artifact

Record the old base digest, replacement base digest, resulting application-image digest, and deployment revision. Check the final image's contents rather than assuming the updated base remains unchanged through later build steps. Additional package installation can reintroduce versions or add components outside the base image.

The running fleet also matters. A registry containing a patched image does not establish that every workload has pulled and started it. Long-running workers, regional deployments, and rollback configurations can retain older images after the main service has been updated.

Treat rebuilding as a tested change

Run the application's relevant checks against the rebuilt image. Changes to system libraries, certificate stores, or language runtimes can affect behavior even when application source is identical. Include a deployment health check and the business function associated with the service.

If the update fails, record the rollback decision and the remaining exposure rather than silently marking the vulnerability resolved. The remediation state should follow the deployed artifact, not the existence of a successful build job.

Verify the running digest

An assessment can select a sample service and trace its declared image to the actual running digest. Compare that digest with the reviewed release and component inventory. Repeat after a rollout and a controlled rollback to identify which evidence remains available for both states.

Record the replacement image, updated workloads, passed checks, and older instances still running. Use that list to track the rollout and plan follow-up work. Pinning a digest makes the chosen image explicit; it still needs an owner who updates it.

Sources

Kubernetes: Images. The web-service update is illustrative, and rollout behavior depends on the deployment platform.

Back to the blogExplore Ceron