malformed metric is produced on the operator's own /metrics page

[redbaron]

Expected behavior (what you expected to happen):

valid metrics are produced on the 9115 port

Actual behavior (what actually happened):

malformed metrics name are produced:

# HELP _PREFIX_kubernetes_client_rate_limiter_latency_seconds {PREFIX}kubernetes_client_rate_limiter_latency_seconds
# TYPE _PREFIX_kubernetes_client_rate_limiter_latency_seconds histogram
_PREFIX_kubernetes_client_rate_limiter_latency_seconds_bucket{url="https://10.100.0.1:443/api/v1/configmaps?fieldSelector=%7Bvalue%7D",verb="GET",le="0.005"} 0

Steps to reproduce:

1. scrape metrics on :9115/metrics

Environment:

• Shell-operator version: 1.14.3
• Kubernetes version: 1.33
• Installation type (kubectl apply, helm chart, etc.):

Anything else we should know?:

Additional information for debugging (if necessary):

hooks metrics on :9115/metrics/hooks seems to be fine

[aspr0d]

I would like to work on this. The kube-client passes metric names with a literal {PREFIX} placeholder through MetricsAdapter without applying app.PrometheusMetricsPrefix. I plan to substitute the prefix in the adapter before registering or updating metrics.