What happened:
Currently inference extension request url := fmt.Sprintf("http://%s/metrics", existing.Address) for metrics, however, in the latest version, pod address is the pod ip and the port is not assigned to the url, so inference extension can not correctly scrape the pod metrics since it always send requests to port 80.

new := &PodMetrics{
		Pod: Pod{
			NamespacedName: types.NamespacedName{
				Name:      pod.Name,
				Namespace: pod.Namespace,
			},
			Address: pod.Status.PodIP,
		},
		Metrics: Metrics{
			ActiveModels: make(map[string]int),
		},
	}

And also, the scrape time is refreshMetricsInterval = flag.Duration("refreshMetricsInterval", 50*time.Millisecond, "interval to refresh metrics via polling pods"). So once the metrics fail to be scraped, the resulting logs can become quite extensive.

What you expected to happen:

How to reproduce it (as minimally and precisely as possible):

Anything else we need to know?:

Environment:

• Kubernetes version (use kubectl version):
• Inference extension version (use git describe --tags --dirty --always):
• Cloud provider or hardware configuration:
• Install tools:
• Others: