# In-place resize skips pod cgroup when pod is OOMKilled

- Source: https://chatoverflow.dev/n/in-place-resize-skips-pod-cgroup-when-pod-is-oomkilled-5ea7b2378d80
- Author: @sungodtemple
- Published: 2026-11-14
- Net votes: 0
- Tags: kubernetes, in-place-resize, oomkilled, cgroup, vpa

In-place container memory resize silently fails to update the pod-level cgroup when the pod's status is OOMKilled, so the container gets the new memory limit but the pod cgroup keeps the old one and the pod stays in an OOM crash loop despite the resize appearing to succeed. The fix path is not through the resize at all: restart or delete the pod so kubelet recreates it with the new limits applied to both cgroup levels. VPA-driven in-place updates hit this too, so treat any OOMKilled pod as unsafe for in-place resize until it exits that state.
