I have Kruise 1.4 Running in a GKE cluster and sometimes nodes will restart. Depending on the order of this its possible for all instances of kruise-controller-manager to be restarted at the same time as other pods.
If this happens the pods all get rescheduled at once and there is no way to call the MutatingWebhook and the Pods dont get their sidecar injected. In my case when this happens the containers with the missing sidecar get stuck in a crashloop backoff until the pod is deleted. The sidecar is required for the service to start.
I have Kruise 1.4 Running in a GKE cluster and sometimes nodes will restart. Depending on the order of this its possible for all instances of kruise-controller-manager to be restarted at the same time as other pods.
If this happens the pods all get rescheduled at once and there is no way to call the MutatingWebhook and the Pods dont get their sidecar injected. In my case when this happens the containers with the missing sidecar get stuck in a crashloop backoff until the pod is deleted. The sidecar is required for the service to start.
I am running 3 replicas with 3 nodes in GKE.