Use Rollout Restart When
You want a controlled restart of Deployments, DaemonSets, or StatefulSets.
You want the controller to handle replacement using the workload's update strategy.
Two ways to refresh running workloads in Kubernetes and why they are not equivalent.
Deleting a pod forces replacement, but it bypasses the intent of the controller. A rollout restart tells the workload to restart cleanly through its normal deployment mechanism.
You want a controlled restart of Deployments, DaemonSets, or StatefulSets.
You want the controller to handle replacement using the workload's update strategy.
You are debugging a stuck pod and want it recreated immediately.
You are dealing with a one-off broken pod rather than refreshing the whole workload.
Restart the workload when it is a workload problem. Delete one pod when it is a pod problem.