ETCD gets OOMKilled when system_reserved is enabled #12566
Description
What would you like to be added
When deploying etcd with etcd_deployment_type: host, it runs under system.slice, which may not provide enough memory for etcd. Please add the ability to configure cgroup and resource limits for etcd separately.
Why is this needed
When system_reserved: true, resources are reserved for system daemons. However, etcd is a Kubernetes component, not a system daemon. This can result in etcd running out of memory and being OOMKilled.