Reverse proxy support?

[neilmfrench]

Hi,
I'm prototyping this with our reverse proxy. However, I'm running into an issue where coredns (exposed) is returning the internal IP of reverse proxy entrypoint, whereas I'd like it to return an external IP so we can have our main DNS delegate to it. Is this currently possible?

Thanks for your help!

[ytsarev]

Hi @neilmfrench , can you elaborate more about the specifics of your setup? What kind of ingress controller do you use and how is it integrated with reverse proxy?

[neilmfrench]

Hi @neilmfrench , can you elaborate more about the specifics of your setup? What kind of ingress controller do you use and how is it integrated with reverse proxy?

Sure, I am using Traefik (other cluster uses nginx, but testing on cluster with Traefik first). Life of a request flows like this

request -> Cloudflare proxy -> external IP (189.XX.XX.XX)-> internal load balancer for Traefik (172.16.XX.XX) -> cluster nodes (not exposed)

I need coredns to respond with the external IP rather than the internal load balancer IP it currently responds with:

Status:
  Healthy Records:
    .....................
      172.16.20.30

I tried creating my own DNSEndpoint with the correct external IP as the target, but it didn't seem to get picked up.

[neilmfrench]

I think the simplest solution would be to allow the reading of the "real" external IP from an annotation on the ingress if it exists, rather than relying only on status.loadBalancer.ingress. Health checks would still use the internal IP, but for building out the local targets list it would use whatever IP the annotation has.

[ytsarev]

@neilmfrench, thanks a lot for the suggestions! That sounds reasonable and should be relatively straightforward to add. We will put it into the roadmap 👍