Metrics and Monitoring

Expose Prometheus metrics from Caddy for monitoring performance, uptime, and operational health

Metrics and Monitoring

Caddy exposes Prometheus metrics for monitoring server health, performance, and operational statistics. Metrics are available through both the admin API and a configurable handler.

Admin Metrics Endpoint

By default, metrics are available on the admin endpoint:

curl http://localhost:2019/metrics

The admin metrics endpoint is enabled by default and requires no configuration.

Metrics Handler

Expose metrics on a public endpoint:

JSON

{
  "apps": {
    "http": {
      "servers": {
        "srv0": {
          "routes": [
            {
              "match": [{"path": ["/metrics"]}],
              "handle": [
                {
                  "handler": "metrics",
                  "disable_openmetrics": false
                }
              ]
            }
          ]
        }
      }
    }
  }
}

Disable OpenMetrics

If your monitoring system doesn’t support OpenMetrics format:

metrics /metrics {
    disable_openmetrics
}

Available Metrics

HTTP Metrics

Request Count

caddy_http_request_count_total{handler="reverse_proxy",server="srv0"}

Request Duration

caddy_http_request_duration_seconds{handler="file_server",server="srv0"}

Request Size

caddy_http_request_size_bytes{handler="reverse_proxy",server="srv0"}

Response Size

caddy_http_response_size_bytes{handler="reverse_proxy",server="srv0"}

Requests In Flight

caddy_http_requests_in_flight{server="srv0"}

Reverse Proxy Metrics

Upstream Request Count

caddy_reverse_proxy_upstreams_request_count_total{upstream="10.0.0.1:8080"}

Upstream Healthy

caddy_reverse_proxy_upstreams_healthy{upstream="10.0.0.1:8080"}

Upstream Request Duration

caddy_reverse_proxy_upstreams_request_duration_seconds{upstream="10.0.0.1:8080"}

TLS Metrics

Handshakes Total

caddy_tls_handshakes_total{conn_policy="0"}

Client Certificates

caddy_tls_client_cert_count{conn_policy="0"}

Admin API Metrics

Request Count

caddy_admin_http_requests_total{handler="config",path="/config/"}

Request Errors

caddy_admin_http_request_errors_total{handler="config",method="POST",path="/config/"}

Prometheus Configuration

Configure Prometheus to scrape Caddy metrics:

prometheus.yml

global:
  scrape_interval: 15s

scrape_configs:
  - job_name: 'caddy'
    static_configs:
      - targets: ['localhost:2019']
    metrics_path: /metrics
    
  # Or from public endpoint
  - job_name: 'caddy-public'
    static_configs:
      - targets: ['metrics.example.com:443']
    scheme: https
    metrics_path: /metrics

Grafana Dashboard

Example Grafana queries:

Request Rate

rate(caddy_http_request_count_total[5m])

Error Rate

sum(rate(caddy_http_request_count_total{status=~"5.."}[5m]))

Response Time (95th percentile)

histogram_quantile(0.95, 
  rate(caddy_http_request_duration_seconds_bucket[5m])
)

Upstream Health

caddy_reverse_proxy_upstreams_healthy

Securing Metrics

Protect your metrics endpoint:

Basic Auth

metrics.example.com {
    basicauth /metrics {
        admin $2a$14$Zkx19XLiW6VYouLHR5NmfOFU0z2GTNmpkT/5qqR7hx4IjWJPDhjvG
    }
    
    metrics /metrics
}

Alert Rules

Example Prometheus alert rules:

groups:
  - name: caddy
    interval: 30s
    rules:
      # High error rate
      - alert: CaddyHighErrorRate
        expr: |
          rate(caddy_http_request_count_total{status=~"5.."}[5m]) > 0.05
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "High 5xx error rate on {{ $labels.server }}"
          
      # Upstream down
      - alert: CaddyUpstreamDown
        expr: caddy_reverse_proxy_upstreams_healthy == 0
        for: 1m
        labels:
          severity: critical
        annotations:
          summary: "Upstream {{ $labels.upstream }} is down"
          
      # High response time
      - alert: CaddyHighLatency
        expr: |
          histogram_quantile(0.95,
            rate(caddy_http_request_duration_seconds_bucket[5m])
          ) > 1.0
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "High response latency on {{ $labels.server }}"
          
      # Certificate expiring
      - alert: CaddyCertificateExpiringSoon
        expr: |
          (caddy_tls_certificate_not_after_timestamp - time()) / 86400 < 7
        labels:
          severity: warning
        annotations:
          summary: "Certificate {{ $labels.subject }} expires in less than 7 days"

Custom Metrics

Register custom metrics in Caddy modules:

import (
    "github.com/prometheus/client_golang/prometheus"
    "github.com/caddyserver/caddy/v2"
)

var myCounter = prometheus.NewCounterVec(
    prometheus.CounterOpts{
        Namespace: "caddy",
        Subsystem: "mymodule",
        Name:      "operations_total",
        Help:      "Total number of operations",
    },
    []string{"type"},
)

func init() {
    caddy.RegisterModule(MyModule{})
}

func (m *MyModule) Provision(ctx caddy.Context) error {
    registry := ctx.GetMetricsRegistry()
    if registry != nil {
        registry.MustRegister(myCounter)
    }
    return nil
}

Complete Example

{
    # Enable metrics on admin endpoint (default)
    admin localhost:2019
}

# Public metrics endpoint with authentication
metrics.internal.example.com {
    # Restrict to internal network
    @metrics {
        path /metrics
        remote_ip 10.0.0.0/8
    }
    
    handle @metrics {
        metrics {
            disable_openmetrics false
        }
    }
    
    handle {
        respond 404
    }
    
    # Optional: Add authentication
    basicauth {
        prometheus $2a$14$...
    }
}

# Main application with instrumentation
app.example.com {
    # All requests are automatically instrumented
    reverse_proxy backend:8080 {
        # Health checks also generate metrics
        health_uri /health
        health_interval 30s
    }
    
    # Log for correlation with metrics
    log {
        output file /var/log/caddy/app.log
        format json
    }
}

Monitoring Stack

Typical monitoring setup:

1. Setup Process

Performance Impact

Metrics collection has minimal performance overhead. Caddy uses efficient Prometheus client libraries with lock-free operations where possible.

Best Practices

1. Metrics Guidelines

Combine metrics with structured logging for complete observability. Use correlation IDs to trace requests across both systems.