charts/vertical-pod-autoscaler ServiceMonitor Support

[ianarsenault]

Feature Request: ServiceMonitor Support

Background

VPA components expose Prometheus metrics on /metrics endpoints, but currently require manual Prometheus configuration
to scrape. ServiceMonitors would enable automatic metric collection in Prometheus Operator environments.

Implementation Requirements

• Add serviceMonitor configuration blocks to values.yaml for all three components
• Create ServiceMonitor template files (recommender, updater, admission-controller)
• Update README.md with ServiceMonitor configuration options
• Add conditional logic to only create ServiceMonitors when enabled

Benefits

• Automatic metric collection for VPA operational monitoring
• Integration with existing Grafana dashboards
• Enable VPA-specific alerting rules for OpsGenie
• Better visibility into VPA performance across customer environments

Acceptance Criteria

• ServiceMonitors are opt-in (disabled by default)
• Support custom labels and scrape intervals
• Follow existing chart patterns and naming conventions
• Include usage examples in documentation

Labels: enhancement, monitoring, prometheus

Priority: medium (nice-to-have for operational visibility)

[ianarsenault]

ServiceMonitors are Kubernetes custom resources used by the Prometheus Operator to automatically configure Prometheus
to scrape metrics from services. Instead of manually configuring Prometheus scrape targets, ServiceMonitors provide a
declarative way to tell Prometheus "scrape this service for metrics."

Why VPA Needs ServiceMonitors

VPA components expose Prometheus metrics on their /metrics endpoints that provide crucial operational insights:

Recommender Metrics:

• vpa_recommender_recommendations_total - Number of recommendations generated
• vpa_recommender_target_resources - Current resource targets per VPA
• vpa_recommender_memory_usage_samples - Memory usage observations

Updater Metrics:

• vpa_updater_evictions_total - Pod eviction counts
• vpa_updater_resource_updates_total - Resource update attempts

Admission Controller Metrics:

• vpa_admission_controller_resource_patches_total - Pod patches applied
• vpa_admission_controller_webhook_duration_seconds - Webhook response times

For Snowplow's multi-tenant infrastructure, these metrics would feed into your Grafana dashboards and OpsGenie alerting
for monitoring VPA health across customer environments.

---

Implementation Details

1. Values.yaml Configuration

Add ServiceMonitor configuration blocks for each component:

  vpa:
    recommender:
      # ... existing config ...
      serviceMonitor:
        # -- Enable ServiceMonitor for Prometheus Operator
        enabled: false
        # -- Additional labels for the ServiceMonitor
        additionalLabels: {}
        # -- ServiceMonitor scrape interval
        interval: 30s
        # -- ServiceMonitor scrape timeout
        scrapeTimeout: 10s
        # -- Metrics path to scrape
        path: /metrics
        # -- Additional ServiceMonitor annotations
        annotations: {}

    updater:
      # ... existing config ...
      serviceMonitor:
        enabled: false
        additionalLabels: {}
        interval: 30s
        scrapeTimeout: 10s
        path: /metrics
        annotations: {}

    admissionController:
      # ... existing config ...
      serviceMonitor:
        enabled: false
        additionalLabels: {}
        interval: 30s
        scrapeTimeout: 10s
        path: /metrics
        annotations: {}