Fix node-exporter port conflict by templatizing hardcoded port 9100

[silokimmo]

Summary

• Replaces 4 hardcoded 9100 port references in the node-exporter DaemonSet template with {{ .Values.services.nodeExporter.metrics }}
• Default port remains 9100 (no change for clusters without conflicts)
• Clusters with a port conflict (e.g. AMD fleet-observability or systemd prometheus-node-exporter occupying 9100) can now override via cluster-values:

apps:
  otel-lgtm-stack:
    helmParameters:
      - name: services.nodeExporter.metrics
        value: "9101"

Background

On nodes where a host-level node_exporter already occupies port 9100 (AMD fleet-observability stack, Ubuntu prometheus-node-exporter systemd package), the cluster-forge DaemonSet pod enters CrashLoopBackOff immediately after bootstrap. The 4 hardcoded 9100 references prevented any per-cluster port override from working.

Affected environments: rck-g03-mi350x, workload-dev/tw016.

Test plan

• Deployed on ephemeral OCI test VM with bloom (fix-node-exporter-port-template branch)
• Replicated port conflict using netcat to reserve port 9100 → confirmed CrashLoopBackOff
• Applied helmParameters override in cluster-values → node-exporter came up Running on port 9101
• Validated on workload-dev (real environment, tw016 had 1974 restarts) → both pods Running after DaemonSet delete + ArgoCD sync

Jira: EAI-6663

🤖 Generated with Claude Code