IssueTracker

Problem you have encountered:

We are using headless service to help resolve the name to ip and vice-a-versa for service pods. We noticed that after 1000 pods, resolution of name to ip stop working. The pod for which dns resolution does not work are the pod/ip which are not listed under headless service endpoints.

What you expected to happen:

name to ip and vice-a-versa should work with headless service infrastructure when there more than 1000 pods under one headless service

Steps to reproduce:

1. Create headless service with label selector
2. Create more than 1000 pod with label selector, use hostname and subdomain parameter in pod spec

** Other information (workarounds you have tried, documentation consulted, etc):**

I tried using cloudDNS as well but still have same issue.

This page document the limit -> https://cloud.google.com/kubernetes-engine/docs/concepts/planning-large-clusters kube-dns supports up to 1000 endpoints per headless service.

We are looking for alternative/workaround for this problem ? using two services is not option at this point.