hello all, I have a question related to sigterm and termination process.
Currently I see that if I send a sigterm to a pod and let it start graceful shoutdown, sometimes (under very heavy load) pod can get new request and since it is under graceful shutdown it will reply as connection refused. what I found, I can avoid this kind of issue with prestop hook and let container wait a little time before start the shutdown and with that give enough time to kubernetes to handle the eviction from endpointslices. For this I need to update all my helm charts to use this prestop. Is there any way to stop sigterm on k8s api level and 1st drain from endpointslices and once it is evicted send the sigterm to pod?