Metrics

Enable Prometheus metrics on any listener. Each listener gets its own isolated registry and scrape endpoint, so different listeners can collect different dimensions.

Configuration

{
  "metrics": {
    "path": "/metrics",
    "collect": {
      "route": true,
      "destination": true,
      "endpoint": false,
      "middleware": true,
      "listener": true
    },
    "histograms": {
      "durationBuckets": [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10],
      "sizeBuckets": [100, 1000, 10000, 100000, 1000000]
    }
  }
}

All fields

Field	Type	Default	Description
`path`	string	`/metrics`	HTTP path for the Prometheus scrape endpoint
`collect`	object	all on except endpoint	Which metric dimensions to collect
`collect.route`	bool	`true`	Route-level metrics (7 metrics)
`collect.destination`	bool	`true`	Destination-level metrics (4 metrics)
`collect.endpoint`	bool	`false`	Per-pod metrics (4 metrics) — high cardinality
`collect.middleware`	bool	`true`	Middleware-level metrics (3 metrics)
`collect.listener`	bool	`true`	Listener-level metrics (3 metrics)
`histograms.durationBuckets`	float[]	Prometheus defaults	Custom boundaries for `_duration_seconds` histograms
`histograms.sizeBuckets`	float[]	Prometheus defaults	Custom boundaries for byte-counting histograms

Examples

All dimensions (except per-pod)

{
  "name": "production",
  "port": 8443,
  "metrics": {
    "path": "/metrics",
    "collect": {
      "route": true,
      "destination": true,
      "endpoint": false,
      "middleware": true,
      "listener": true
    }
  }
}

This is the recommended configuration. 21 metrics with manageable cardinality.

Including per-pod metrics

{
  "name": "debug",
  "port": 9090,
  "metrics": {
    "path": "/metrics",
    "collect": {
      "route": true,
      "destination": true,
      "endpoint": true,
      "middleware": true,
      "listener": true
    }
  }
}

All 22 metrics. Enable endpoint only when you need to debug individual pods — it generates one time series per pod × status code combination.

Minimal (route metrics only)

{
  "name": "minimal",
  "port": 8080,
  "metrics": {
    "path": "/metrics",
    "collect": {
      "route": true,
      "destination": false,
      "endpoint": false,
      "middleware": false,
      "listener": false
    }
  }
}

Custom histogram buckets for low-latency APIs

{
  "name": "fast-api",
  "port": 8080,
  "metrics": {
    "path": "/metrics",
    "histograms": {
      "durationBuckets": [0.001, 0.002, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1]
    }
  }
}

Custom metrics path

{
  "metrics": {
    "path": "/-/prometheus/metrics"
  }
}

Two listeners, different metrics

[
  {
    "name": "public",
    "port": 443,
    "metrics": {
      "path": "/metrics",
      "collect": {"route": true, "destination": true, "middleware": true, "listener": true}
    }
  },
  {
    "name": "internal",
    "port": 8080,
    "metrics": {
      "path": "/metrics",
      "collect": {"route": true, "destination": true, "endpoint": true, "middleware": true, "listener": true}
    }
  }
]

No metrics

Omit the metrics field entirely. No Prometheus endpoint is served on this listener.

Available metrics

Route metrics

Enabled by default (collect.route: true).

Metric	Type	Labels
`vrata_route_requests_total`	counter	route, group, method, status_code, status_class
`vrata_route_duration_seconds`	histogram	route, group, method
`vrata_route_request_bytes_total`	counter	route, group
`vrata_route_response_bytes_total`	counter	route, group
`vrata_route_inflight_requests`	gauge	route, group
`vrata_route_retries_total`	counter	route, group, attempt
`vrata_mirror_requests_total`	counter	route, destination

# Error rate per route
sum(rate(vrata_route_requests_total{status_class="5xx"}[5m])) by (route)
/ sum(rate(vrata_route_requests_total[5m])) by (route)

# P99 latency
histogram_quantile(0.99, sum(rate(vrata_route_duration_seconds_bucket[5m])) by (route, le))

# Retry storm detection
sum(rate(vrata_route_retries_total[5m])) by (route)

Destination metrics

Enabled by default (collect.destination: true).

Metric	Type	Labels
`vrata_destination_requests_total`	counter	destination, status_code, status_class
`vrata_destination_duration_seconds`	histogram	destination
`vrata_destination_inflight_requests`	gauge	destination
`vrata_destination_circuit_breaker_state`	gauge	destination

Circuit breaker state: 0 = closed, 1 = open, 2 = half-open.

# Canary error rate comparison
sum(rate(vrata_destination_requests_total{status_class="5xx", destination="canary"}[5m]))
/ sum(rate(vrata_destination_requests_total{destination="canary"}[5m]))

# Circuit breaker firing
vrata_destination_circuit_breaker_state == 1

Endpoint metrics

Off by default (collect.endpoint: false) — high cardinality. Enable when you need per-pod visibility.

Metric	Type	Labels
`vrata_endpoint_requests_total`	counter	destination, endpoint, status_code, status_class
`vrata_endpoint_duration_seconds`	histogram	destination, endpoint
`vrata_endpoint_healthy`	gauge	destination, endpoint
`vrata_endpoint_consecutive_5xx`	gauge	destination, endpoint

With 100 destinations × 10 pods × 5 status codes = 5000 time series just for vrata_endpoint_requests_total.

# Find the slow pod
histogram_quantile(0.99, rate(vrata_endpoint_duration_seconds_bucket{destination="api"}[5m])) by (endpoint)

# Unhealthy endpoints
vrata_endpoint_healthy == 0

Middleware metrics

Enabled by default (collect.middleware: true).

Metric	Type	Labels
`vrata_middleware_duration_seconds`	histogram	middleware, type
`vrata_middleware_rejections_total`	counter	middleware, type, status_code
`vrata_middleware_passed_total`	counter	middleware, type

# JWT validation overhead
histogram_quantile(0.99, rate(vrata_middleware_duration_seconds_bucket{type="jwt"}[5m]))

# Rate limit effectiveness
rate(vrata_middleware_rejections_total{type="rateLimit", status_code="429"}[5m])

# Auth denial rate
sum(rate(vrata_middleware_rejections_total{type="jwt"}[5m]))
/ (sum(rate(vrata_middleware_rejections_total{type="jwt"}[5m])) + sum(rate(vrata_middleware_passed_total{type="jwt"}[5m])))

Listener connection metrics

Enabled by default (collect.listener: true).

Metric	Type	Labels
`vrata_listener_connections_total`	counter	listener, address
`vrata_listener_active_connections`	gauge	listener, address
`vrata_listener_tls_handshake_errors_total`	counter	listener, address

# Connection rate
rate(vrata_listener_connections_total[5m])

# Certificate problems
rate(vrata_listener_tls_handshake_errors_total[5m])

# Connection saturation
vrata_listener_active_connections