Traefik routes to another service instead of returning 503 when /api backend is down

Traefik Hub
1

Hi everyone,

I have configured multiple Traefik Ingress resources for the same host: one handling the / path prefix and another handling /api. Each Ingress points to a different service:

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: a
spec:
ingressClassName: traefik
rules:

  • host: "example.com"
    http:
    paths:
  • path: /api
    pathType: Prefix
    backend:
    service:
    name: a
    port:
    number: 9090

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: b
spec:
ingressClassName: traefik
rules:

  • host: "example.com"
    http:
    paths:
  • path: /
    pathType: Prefix
    backend:
    service:
    name: b
    port:
    number: 8080

When the pod behind service a is running properly, requests to /api are handled as expected. However, when that pod becomes unavailable or unhealthy, requests to /api/... are routed to service b (the / path) instead of returning an error.

As far as I understand, this might be default behavior, but I would prefer Traefik to return a 503 Service Unavailable response (or a similar error) when the /api backend is not available, rather than falling back to another service.

Is there a way to configure Traefik to achieve this behavior?

2

For the Docker provider there is allowEmptyServices, maybe there is something similar for k8s.

3

Hi,

you could try merging the two ingresses into one (though, I'm not 100% if it'll fix it, but I don't see a reason to have two different ingress objects for that use case). I'm a bit low on time and unable to setup such a scenario to test it.
The same question was asked a couple of days ago here (there, you'll also find the two merged ingress objects):