Resize a running Kubernetes Pod's CPU/memory requests and limits in place without restarting it

Steps

1. Confirm the cluster is on Kubernetes v1.35 or later, where in-place pod resize (InPlacePodVerticalScaling) is stable and enabled by default; on older supported versions the feature gate must be enabled explicitly on kube-apiserver and kubelet.
2. Set resizePolicy per container in the Pod spec (e.g. cpu: NotRequired, memory: RestartContainer) to control whether a change to that resource triggers a container restart.
3. Patch the pod's resize subresource with a sufficiently new kubectl client (v1.32+), e.g. kubectl patch pod <name> --subresource resize --patch '{"spec":{"containers":[{"name":"<container>","resources":{"requests":{...},"limits":{...}}}]}}'.
4. Watch the pod's resize status and container restart counts to confirm whether the change applied without a restart or triggered a restart per its resizePolicy.
5. For memory limit decreases, expect the kubelet to only apply the resize once current usage is already below the new limit; a resize can be deferred and retried automatically as usage drops or node capacity frees up.