In our production environment, we often encounter the need to modify the CPU or memory resources of a pod. However, making these changes always triggers a pod restart, which is unacceptable for some applications. Could the community support a feature that allows resource modification without restarting the pod? Is there a specific reason why this feature has not been implemented? Thank you.
Why? Because it is much harder than it sounds.
It’s already in alpha, though
This is really good news.
Before the in-place resource update feature is ready, maybe you want to recheck your workloads and see if the requirements are reasonable. Pods, by design, are ephemeral objects. That is, they are designed to be short-lived. From an end user’s perspective, you will focus more on the exposed service. As long as the service is available, you won’t care whether a Pod is restarted. Transparent Pod restarts are sometimes a good thing.
There are for sure some valid cases where in-place resource updates are desirable. What I’m suggesting is that we can spend more time on the “unacceptable” part, before jumping onto some fancy (maybe unstable) features.