Manage container images

Prune old container images in the private registry to reclaim space: remove untagged images older than 30 days and images for project-legacy that have not been pulled in 90 days. Run the cleanup on registry-01 during the 03:00 maintenance window and ensure the garbage collector runs after deletion. Before deleting, snapshot registry metadata and notify the projects team with a 48-hour warning.

Before I run the image cleanup, make the change easy to approve for platform stakeholders: show the projected disk space reclaimed, list the exact image names and tags that will be deleted, surface any images used by running deployments, and flag images that are part of active CI pipelines. Include a one-click rollback plan to re-push critical images and a short note on how to avoid accidental future deletions.

I cleaned untagged images older than 30 days and removed stale project-legacy tags from registry-01; an hourly job failed because it pulled tag v3.2.1 which no longer exists. I don’t know whether that job could have been sampling latest or an immutable CI cache was missed. The deployment is business-critical for an analytics dashboard used by Sales. What immediate action gets the dashboard back while preserving the cleanup objective?

Every quarter we run registry pruning and then encounter a critical pipeline or production service that relied on an old tag we deleted, costing hours to recover. Where are we most likely failing: ownership of images, lack of retention policy, poor tagging practices, or missing registry protection? Recommend one habit change to prevent these incidents and one quick metric to confirm fewer accidental deletions.