Incorrect service name emitted in some metrics

[adrianmoisey]

What happened:

When defining two backends, as such:

     - backend:
          service:
            name: service-a
            port:
              number: 80
        path: /
        pathType: Exact
      - backend:
          service:
            name: service-b
            port:
              number: 80
        path: /
        pathType: Prefix

We only get metrics for the service-a backend.

What you expected to happen:

To be able to get metrics for each backend.

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

NGINX Ingress controller
  Release:       v1.3.0
  Build:         2b7b74854d90ad9b4b96a5011b9e8b67d20bfb8f
  Repository:    https://github.com/kubernetes/ingress-nginx
  nginx version: nginx/1.19.10

Kubernetes version (use kubectl version):

Server Version: version.Info{Major:"1", Minor:"25", GitVersion:"v1.25.8", GitCommit:"0ce7342c984110dfc93657d64df5dc3b2c0d1fe9", GitTreeState:"clean", BuildDate:"2023-03-15T13:33:02Z", GoVersion:"go1.19.7", Compiler:"gc", Platform:"linux/amd64"}

Environment:

• Cloud provider or hardware configuration: AWS (via kOps)

• OS (e.g. from /etc/os-release): Ubuntu 20.04.5 LTS

• Kernel (e.g. uname -a): 5.15.0-1031-aws

• Install tools: kOps

  • Please mention how/where was the cluster created like kubeadm/kops/minikube/kind etc.

• Basic cluster related info:

  • kubectl version Server Version: version.Info{Major:"1", Minor:"25", GitVersion:"v1.25.8", GitCommit:"0ce7342c984110dfc93657d64df5dc3b2c0d1fe9", GitTreeState:"clean", BuildDate:"2023-03-15T13:33:02Z", GoVersion:"go1.19.7", Compiler:"gc", Platform:"linux/amd64"}

  • kubectl get nodes -o wide Ubuntu 20.04.5 LTS 5.15.0-1031-aws containerd://1.6.18

• How was the ingress-nginx-controller installed:
  Installed via Helm using ArgoCD

• Current State of the controller:
  Controller is running and working as expected, except for this metric

• Current state of ingress object, if applicable:
  All Ingress objects are fine and the controller handles requests fine.

How to reproduce this issue:
Create an ingress object with multiple / paths, of different types.
Send traffic to both matches / and /test.
Look at resulting metrics.

Anything else we need to know:

I looked at the resulting configuration and compared them to see if I can find the issue. This seems to highlight the bug:

$ diff -u a b
--- a	2023-07-13 10:56:34
+++ b	2023-07-13 10:56:42
@@ -1,4 +1,4 @@
-		location / {
+		location = / {

 			set $namespace      "mynamespace";
 			set $ingress_name   "myingress";
@@ -46,7 +46,7 @@
 			port_in_redirect off;

 			set $balancer_ewma_score -1;
-			set $proxy_upstream_name "mynamespace-service-a-80";
+			set $proxy_upstream_name "mynamespace-service-b-80";
 			set $proxy_host          $proxy_upstream_name;
 			set $pass_access_scheme  $scheme;

This config is fine, except that set $service_name is the same in both.

It seems that the problem is happening around here:

• https://github.com/kubernetes/ingress-nginx/blob/controller-v1.3.0/rootfs/etc/nginx/template/nginx.tmpl#L1246
• https://github.com/kubernetes/ingress-nginx/blob/controller-v1.3.0/rootfs/etc/nginx/template/nginx.tmpl#L1189

Which I think comes from https://github.com/kubernetes/ingress-nginx/blob/controller-v1.3.0/internal/ingress/controller/template/template.go#L1061-L1149

I think the expected diff is as such:

$ diff -u a b
--- a	2023-07-18 13:13:39
+++ b	2023-07-13 10:56:42
@@ -1,8 +1,8 @@
-		location / {
+		location = / {

 			set $namespace      "mynamespace";
 			set $ingress_name   "myingress";
-			set $service_name   "service-a";
+			set $service_name   "service-b";
 			set $service_port   "80";
 			set $location_path  "/";
 			set $global_rate_limit_exceeding n;
@@ -46,7 +46,7 @@
 			port_in_redirect off;

 			set $balancer_ewma_score -1;
-			set $proxy_upstream_name "mynamespace-service-a-80";
+			set $proxy_upstream_name "mynamespace-service-b-80";
 			set $proxy_host          $proxy_upstream_name;
 			set $pass_access_scheme  $scheme;

[k8s-ci-robot]

This issue is currently awaiting triage.

If Ingress contributors determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

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/test-infra repository.

[longwuyuan]

/remove-kind bug

Does this seem like your use case https://kubernetes.github.io/ingress-nginx/user-guide/monitoring/#wildcard-ingresses

[adrianmoisey]

I don't think so. We do use the --metrics-per-host=false option
The particular tag we're worried about is service, which is the backend service that is bring routed to, not the host

[strongjz]

Can you test with /test as a specific path and see if the issue is the same? There may be an issue with pathType and the same path, /, the controller may not be differianating them.

Also can you try this in the latest relase, 1.8.1?

[github-actions]

This is stale, but we won't close it automatically, just bare in mind the maintainers may be busy with other tasks and will reach your issue ASAP. If you have any question or request to prioritize this, please reach #ingress-nginx-dev on Kubernetes Slack.

[adrianmoisey]

Can you test with /test as a specific path and see if the issue is the same? There may be an issue with pathType and the same path, /, the controller may not be differianating them.

Adding a /test path that is an exact match works as expected.
I think the problem is exactly as you say, the controller isn't differentiating between them.

Also can you try this in the latest relase, 1.8.1?

I've tested with 1.10.0 and still get this issue

[adrianmoisey]

I think I understand what is happening here. I plan to submit a PR if that is fine with you

[longwuyuan]

/assign