Is there a way of avoiding downtime when updating a Kubernetes LB Service's ports on GKE?
We run a platform on GKE that exposes user applications through a Load Balancer Service.
When the Service's ports are updated (add or delete), we find that HTTP and TCP connections to the Load Balancer time out for a few seconds.
If we can't avoid the disruption, we can pre-allocate a set of ports on the LB Service so that it doesn't need to be updated. Would prefer to keep the port allocation dynamic, however.