This is not a good idea, you're doubling the load on conntrack. Maybe ingress-nginx should fix their config generation instead. Had a cluster with at least one

This is not a good idea, you're doubling the load on conntrack. Maybe ingress-nginx should fix their config generation instead. Had a cluster with at least one change per second to the ingresses, and nginx would regularly just die (and orphan existing requests after 20 seconds).