In rare cases, a cluster control plane upgrade can cause an Autopilot node to

## Issue Issue In rare cases, a cluster control plane upgrade can cause an Autopilot node to enter into a state in which new system and user Pods are unable to run, which then causes issues such as broken Pod networking. GKE is regularly detecting this issue, and when possible GKE is mitigating new occurrences of this issue. For more details, see[Pods unable to run on a Node due to NRI RunPodSandbox failed](https://docs.cloud.google.com/kubernetes-engine/docs/troubleshooting/autopilot-clusters#nri-runpodsandbox-error), and to mitigate this issue yourself, follow the procedure in the section[Consistently unreliable workload performance on a specific node](https://docs.cloud.google.com/kubernetes-engine/docs/troubleshooting/autopilot-clusters#unreliable-workloads-specific-node).