Prometheus PodMonitor for Knative is incorrectly generated

[Evynglais]

What happened?

Using the Prometheus trait with a Knative Camel-K integration results in an invalid PodMonitor configuration (a port that doesn't exist).

Steps to reproduce

1. Deploy a Camel-K integration as a Knative service, with --trait prometheus.enabled=true.

kamel run "${SCRIPT_DIR}"/pre.ingress.camel.yaml --name pre-ingress \
    --trait knative-service.min-scale=1 \
    --trait prometheus.enabled=true \
    --property quarkus.kubernetes-config.enabled=true \
    --property quarkus.kubernetes-config.config-maps=pre.ingress.ibm.config

2. Observe the target group on Prometheus; it will be empty.
   
3. Observe the podmonitor created by Camel-K; it will reference port 'h2c' which doesn't exist.
   

Actual ports:

Relevant log output

No response

Camel K version

2.5.0

[squakez]

That is something expected according to Knative spec: https://github.com/knative/specs/blob/main/specs/serving/runtime-contract.md#protocols-and-ports - you may need to verify on the KnativeService custom resource if that is configured correctly. Alternatively you may need to specify the port name on the container trait. @hernanDatgDev @mertdotcc I think you folks are using Knative + Prometheus, would you mind to confirm if that is the expected behavior?

[Evynglais]

The way that camel is working as a Knative Service is expected and working fine; the PodMonitor looks to me that it is generated incorrectly.

[squakez]

I see this is a possible regression introduced by #1555. The "problem" is that the KnativeService resource is setting the port as

 - containerPort: 8080
   name: h2c
   protocol: TCP

However the Deployment (hence, the Pod) is set as:

- containerPort: 8080
  name: user-port
  protocol: TCP

As a most immediate workaround you can try to put a name on the port using Container portName trait configuration. Can you verify if that one temporarily solves the problems?