In-Place Pod Resize in K8s 1.33: Latency and Failure Modes
Explore the latency and failure modes of in-place pod resize in K8s 1.33.
K8s 1.33 has graduated the pods/resize subresource to GA, allowing memory limits to be changed without restarting containers. However, the actual resize duration can take 1 to 5 seconds, with longer delays under heavy node load. Additionally, simply raising limits without adjusting requests can lead to scheduler overcommitment, making it crucial to manage both parameters together.
This synthesis was produced from its source by AI; there is no human editor or manual review step. How we work