A Deployment manages a Pod template and a desired replica count. It uses ReplicaSets to keep the requested number of matching Pods available. Work on the controller when the team needs a repeatable application service; changing or deleting one managed Pod doesn't change the desired workload. Scaling changes capacity without changing the image. Read the target resource name carefully when several previews exist in the same namespace. A successful scale command against the wrong Deployment leaves the actual service unchanged. These tasks vary the requested count and preserve the image, then ask you to reason about what the controller will recreate.