Nginx-controller re-loads its config when TCP proxied endpoints change

[troodes]

What happened:
We use the ingress-nginx to proxy several TCP ports, as well as HTTP services. By "TCP proxied ports", I mean this: https://github.com/kubernetes/ingress-nginx/blob/main/docs/user-guide/exposing-tcp-udp-services.md

I'm noticing nginx config reloads happen when endpoints of the proxied TCP services change, but no reloads occur when HTTP proxied endpoints change. Based on https://kubernetes.github.io/ingress-nginx/how-it-works/ I would expect reloads to not happen on endpoints status changes, but that doesn't appear to be the case for the TCP proxied endpoints.

What you expected to happen:

Ideally, no reloads would happen, similar to the way it is handled for HTTP proxied service endpoints.

NGINX Ingress controller version (exec into the pod and run nginx-ingress-controller --version.):

NGINX Ingress controller
  Release:       v1.2.0
  Build:         a2514768cd282c41f39ab06bda17efefc4bd233a
  Repository:    https://github.com/kubernetes/ingress-nginx
  nginx version: nginx/1.19.10
-------------------------------------------------------------------------------

Kubernetes version (use kubectl version):
kubectl version: 1.23.7

Environment:

• baremetal, Red Hat Enterprise Linux release 8.6

• How was the ingress-nginx-controller installed:
  # helm ls -A | grep -i ingress nginx-ingress default 3 2022-11-30 14:18:32.16446935 -0500 EST deployed ingress-nginx-4.1.2 1.2.0

How to reproduce this issue:
Expose a TCP port, as described here: https://github.com/kubernetes/ingress-nginx/blob/main/docs/user-guide/exposing-tcp-udp-services.md
For a proxied TCP service, restart the pod of the app associated with it
Observe the following in the nginx logs:

I1205 19:23:12.380089       7 controller.go:166] "Configuration changes detected, backend reload required"
I1205 19:23:12.552513       7 controller.go:183] "Backend successfully reloaded"

I'm seeing 2x reloads happen when the TCP endpoint goes down, and then comes back up presumably?
In my case, frequent reloads are leading to OOM problems, but at this point I'm just trying to figure out if re-loads should be expected for TCP proxied ports?

[longwuyuan]

@troodes, the connection from the controller pod to a new endpoint has to be re-established, I am assuming.

/remove-kind bug
/kind support

[strongjz]

/assign @rikatz
/triage accepted
/priority backlog

We need to review if the lua code that is responsible for the hot reload on HTTP but not TCP so we need to investigate that.

[k8s-triage-robot]

This issue has not been updated in over 1 year, and should be re-triaged.

You can:

• Confirm that this issue is still relevant with /triage accepted (org members only)
• Close this issue with /close

For more details on the triage process, see https://www.kubernetes.dev/docs/guide/issue-triage/

/remove-triage accepted

[longwuyuan]

The tcp/udp proxy feature is being deprecated #11666, so this issue can not track any action items hence closing it.

/close

[k8s-ci-robot]

@longwuyuan: Closing this issue.

In response to this:

The tcp/udp proxy feature is being deprecated #11666, so this issue can not track any action items hence closing it.

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.