Newt Helm chart requires manual `NEWT_METRICS_PROMETHEUS_ENABLED=true` and non-loopback `adminAddr` for Prometheus scraping

[slydlake]

Summary

In my Kubernetes setup (Flux + Prometheus Operator), Newt metrics only work when we manually set both:

• extraEnv.NEWT_METRICS_PROMETHEUS_ENABLED: "true"
• global.metrics.adminAddr: "0.0.0.0:9090"

Without these overrides, ServiceMonitor scraping fails.

Environment

• Helm chart: fosrl/newt (1.2.0)
• Deployment method: Flux HelmRelease
• Monitoring: Prometheus Operator (ServiceMonitor)
• Metrics service enabled via chart values (global.metrics.service.enabled: true)
• ServiceMonitor enabled via chart values (global.metrics.serviceMonitor.enabled: true)

What I observe

• Metrics Service and Endpoints are created and point to port 9090.
• ServiceMonitor scrapes the metrics service port.
• Newt pod does not expose working metrics unless we set both values above.

Observed failure modes:

• connection refused on podIP:9090 when Newt is still bound to loopback/default admin address.
• 404 /metrics when Prometheus exporter is not explicitly enabled.

Why both settings are required in my environment

1) NEWT_METRICS_PROMETHEUS_ENABLED: "true"

global.metrics.enabled: true in chart values does not reliably enable the Newt Prometheus exporter at runtime by itself.

In our rendered deployment:

• NEWT_METRICS_PROMETHEUS_ENABLED is not present by default.
• --metrics CLI flag is also not present.

Result: /metrics is not served (404), even if admin HTTP is reachable.

2) adminAddr: "0.0.0.0:9090"

Default admin binding is loopback-oriented (127.0.0.1:2112 behavior path), which is not reachable through Kubernetes Service/Endpoints scraping from outside the container namespace.

My ServiceMonitor path is:
Prometheus -> Service -> PodIP:9090

Therefore Newt must bind to a non-loopback address and matching port.
0.0.0.0:9090 satisfies this and aligns with the chart’s metrics service defaults.

Important: ":9090" is not accepted by chart schema (requires host:port format), so 0.0.0.0:9090 is required.

Current workaround

values:
  global:
    extraEnv:
      NEWT_METRICS_PROMETHEUS_ENABLED: "true"
    metrics:
      enabled: true
      adminAddr: "0.0.0.0:9090"
      service:
        enabled: true
      serviceMonitor:
        enabled: true