[BUG] Kubernetes service status check fails when ingress name differs from service name

[Mik4sa]

Description

I'm running youtrack in my kubernetes cluster but homepage fails for the online / status check.

These are my two needed ingress configs:

---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: youtrack-event-source-bus
  namespace: youtrack-prod
  annotations:
    nginx.ingress.kubernetes.io/force-ssl-redirect: "true"
    nginx.ingress.kubernetes.io/proxy-read-timeout: "86400"
    nginx.ingress.kubernetes.io/proxy-send-timeout: "86400"
    nginx.ingress.kubernetes.io/configuration-snippet: |
      proxy_cache off;
      proxy_set_header Connection '';
      chunked_transfer_encoding off;
spec:
  ingressClassName: nginx
  rules:
    - host: youtrack.svc.local
      http:
        paths:
        - backend:
            service:
              name: youtrack
              port:
                number: 8080
          pathType: Prefix
          path: /api/eventSourceBus
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: youtrack-root
  namespace: youtrack-prod
  annotations:
    nginx.ingress.kubernetes.io/force-ssl-redirect: "true"
    nginx.ingress.kubernetes.io/proxy-buffering: "on"
    nginx.ingress.kubernetes.io/proxy-buffers-number: "8"
    nginx.ingress.kubernetes.io/proxy-buffers: "64k"
    nginx.ingress.kubernetes.io/proxy-busy-buffers-size: "128k"
    nginx.ingress.kubernetes.io/configuration-snippet: |
      proxy_busy_buffers_size 128k;
    nginx.ingress.kubernetes.io/proxy-buffer-size: "64k"
    nginx.ingress.kubernetes.io/proxy-body-size: "10m"
    gethomepage.dev/enabled: "true"
    gethomepage.dev/name: "YouTrack"
    gethomepage.dev/icon: "/images/youtrack.svg"
spec:
  ingressClassName: nginx
  rules:
    - host: youtrack.svc.local
      http:
        paths:
        - backend:
            service:
              name: youtrack
              port:
                number: 8080
          pathType: Prefix
          path: /

The request url from homepage is: https://dashboard.svc.local/api/kubernetes/status/youtrack-prod/youtrack-root? this results into: {"status":"not found"}
When making a custom request to: https://[dashboard.svc.local/api/kubernetes/status/youtrack-prod/youtrack](https://dashboard.svc.local/api/kubernetes/status/youtrack-prod/youtrack) everything is working. The result then is: {"status":"running"}

Looks like the name of the ingress resources is used and not the actual referenced service name.

homepage version

v0.9.5 (eccd96a)

Installation method

Other (please describe above)

Configuration

No response

Container Logs

No response

Browser Logs

No response

Troubleshooting

See above for all details.

[shamoon]

There has been tons of back-and-forth about this with our k8s users, so I don’t think we’re going to change things here. as far as I can tell things aren’t so standardized in k8s so what one person calls a ‘bug’ is fine for others.

I don’t use k8s, so that also makes this difficult to be definitive about, but perhaps look through old discussions about this.

[Mik4sa]

Would it then be possible to have a setting to control this behavior or a setting to change the actual url used for that?

[shamoon]

Take a look at #2182 and #2701

[Mik4sa]

Thanks, I fixed it with setting the annotation gethomepage.dev/app: "youtrack"

[blackliner]

I think the problem here is the authors knowledge about how ingress works. See this comment:

homepage/docs/configs/kubernetes.md

Line 118 in b9b7c48

gethomepage.dev/app: emby-app # optional, may be needed if app.kubernetes.io/name != ingress metadata.name

It makes the impression that the ingress name has to be equal to the label selector, but that is rarely the case. The correct approach would be to find the pods the same way kubernetes ingress does it: by looking at the service that is mentioned in the ingress! Right now the logic is flawed, as it misses this relevant step between ingress and pods.

I would really consider adding this feature, thanks

[vjm]

I think you are right -- however, it looks like one way to solve this right now is to use the gethomepage.dev/pod-selector annotation. for isolated namespaces, you can use a simple annotation like gethomepage.dev/pod-selector: "". This works well for things like Longhorn.

For something in a mixed namespace, you can be more specific. For example, running the weaveworks/weave-gitops flux dashboard in the flux-system namespace, you will instead want gethomepage.dev/pod-selector: "app.kubernetes.io/name=weave-gitops".

Documentation for this is light but can be found here

Hope this helps!

[mrsimonemms]

Can confirm that @vjm's solution works. I'm going to update the docs and raise a PR

[mrsimonemms]

#4316