golib/prometheus

Prometheus Package

Production-ready Prometheus metrics integration for Go applications with Gin web framework support, dynamic metric management, and thread-safe operations.

---

Table of Contents

• Overview
• Key Features
• Installation
• Architecture
• Quick Start
• Performance
• Use Cases
• Subpackages
  • metrics - Metric Definitions
  • pool - Metric Pool Management
  • types - Core Type Definitions
  • webmetrics - Pre-configured Web Metrics
  • bloom - Bloom Filter Implementation
• Best Practices
• Testing
• Contributing
• Future Enhancements
• License

---

Overview

This library provides production-ready Prometheus metrics integration for Go applications, with first-class support for the Gin web framework. It emphasizes dynamic metric management, thread safety, and efficient metric collection with minimal overhead.

Design Philosophy

1. Dynamic Management: Register, unregister, and update metrics at runtime
2. Thread-Safe: Atomic operations and mutex protection for concurrent operations
3. Gin-First: Seamless integration with Gin middleware and routing
4. Type-Safe: Strong typing for all metric types (Counter, Gauge, Histogram, Summary)
5. Modular: Independent subpackages that work together seamlessly
6. Performance: Minimal overhead with efficient metric collection

---

Key Features

• Dynamic Metric Management: Add, remove, and update metrics at runtime without restarts
• Thread-Safe Operations: Atomic state management and synchronized metric collection
• Gin Integration: Built-in middleware, route handlers, and context support
• Multiple Metric Types:
  • Counter: Cumulative values (requests, errors, bytes transferred)
  • Gauge: Values that can increase or decrease (active connections, queue size)
  • Histogram: Distributions with buckets (request latency, response size)
  • Summary: Streaming quantiles (precise percentiles)
• Path Exclusion: Exclude specific paths from monitoring (e.g., health checks)
• Pre-configured Metrics: 11 production-ready web metrics via webmetrics subpackage
• Bloom Filter Support: Efficient unique value tracking (e.g., unique IP addresses)
• Standard Interfaces: Compatible with standard prometheus/client_golang

---

Installation

go get github.com/nabbar/golib/prometheus

---

Architecture

Package Structure

The package is organized into specialized subpackages:

prometheus/
├── metrics/             # Metric definitions and collectors
├── pool/                # Thread-safe metric pool management
├── types/               # Core type definitions and interfaces
├── webmetrics/          # Pre-configured web server metrics
├── bloom/               # Bloom filter for unique tracking
├── interface.go         # Main Prometheus interface
├── model.go             # Implementation of Prometheus interface
├── pool.go              # Metric pool operations
└── route.go             # Gin route handlers and middleware

Component Overview

┌─────────────────────────────────────────────────────────┐
│                  Prometheus Package                      │
│  New(), Middleware(), Expose(), Metric Management       │
└──────────────┬──────────────┬──────────────┬────────────┘
               │              │              │
      ┌────────▼─────┐  ┌────▼─────┐  ┌────▼──────┐
      │   metrics    │  │   pool   │  │   types   │
      │              │  │          │  │           │
      │  Counter     │  │ Add/Get  │  │ Metric    │
      │  Gauge       │  │ Del/List │  │ MetricType│
      │  Histogram   │  │ Walk     │  │ Register  │
      │  Summary     │  │          │  │           │
      └──────────────┘  └──────────┘  └───────────┘
               │
      ┌────────▼──────────┐
      │   webmetrics      │
      │                   │
      │  11 Pre-configured│
      │  Production Metrics│
      └───────────────────┘

Component	Purpose	Thread-Safe	Coverage
metrics	Metric definition and collection	✅	95.5%
pool	Metric pool management	✅	72.5%
types	Type definitions and registration	✅	100%
webmetrics	Pre-configured web metrics	✅	0% (simple constructors)
bloom	Bloom filter for unique tracking	✅	94.7%
Root	Main interface and middleware	✅	90.9%

Metric Flow

┌─────────────┐         ┌──────────────┐         ┌─────────────┐
│ HTTP Request│────────▶│  Middleware  │────────▶│   Handler   │
└─────────────┘         └──────────────┘         └─────────────┘
                               │                         │
                               │ (set start time)        │
                               ▼                         │
                        ┌──────────────┐                 │
                        │ Gin Context  │                 │
                        │ - Start Time │                 │
                        │ - Path Info  │                 │
                        └──────────────┘                 │
                               │                         │
                               │◀────────────────────────┘
                               │ (after handler)
                               ▼
                        ┌──────────────┐
                        │   Collect    │
                        │   Metrics    │
                        └──────────────┘
                               │
                               ▼
                        ┌──────────────┐
                        │  Prometheus  │
                        │   Registry   │
                        └──────────────┘
                               │
                               ▼
                        ┌──────────────┐
                        │ /metrics     │
                        │  Endpoint    │
                        └──────────────┘

---

Performance

Memory Efficiency

• Metric Storage: ~200 bytes per counter/gauge, ~2KB per histogram (with 10 buckets)
• Pool Overhead: O(n) where n is number of registered metrics
• Concurrent Safe: Lock-free reads, synchronized writes

Thread Safety

All operations are thread-safe through:

• Atomic Operations: atomic.Int32 for scalar values
• Value Protection: libatm.Value for complex types (slices, maps)
• Mutex-Free Reads: Lock-free metric access
• Concurrent Collection: Parallel metric collection with semaphore control

Throughput Benchmarks

Operation	Overhead	Memory	Notes
Counter Inc	~50ns	O(1)	Atomic operation
Histogram Observe	~200ns	O(1)	Bucket lookup
Metric Collection	~500ns	O(1)	Per metric
Middleware	~1µs	O(1)	Per request

For 1000 req/s: ~0.1% CPU overhead

---

Use Cases

This library is designed for scenarios requiring Prometheus monitoring:

Web APIs

• Monitor request rates, latencies, and error rates
• Track endpoint performance and SLA compliance
• Identify slow endpoints and bottlenecks
• Monitor active connections and resource usage

Microservices

• Service-level metrics with custom prefixes
• Cross-service performance comparison
• Error tracking and alerting
• Distributed tracing correlation

Production Monitoring

• Real-time dashboards with Grafana
• Automated alerting based on thresholds
• Capacity planning with historical data
• Incident investigation and debugging

Performance Testing

• Load testing metrics and analysis
• Latency distribution analysis
• Identify performance regressions
• Benchmark different configurations

---

Quick Start

Basic Setup

package main

import (
    "github.com/gin-gonic/gin"
    libctx "github.com/nabbar/golib/context"
    "github.com/nabbar/golib/prometheus"
    "github.com/nabbar/golib/prometheus/webmetrics"
)

func main() {
    // Create Prometheus instance
    prm := prometheus.New(libctx.NewContext)
    
    // Configure slow request threshold (5 seconds)
    prm.SetSlowTime(5)
    
    // Register essential metrics
    prm.AddMetric(true, webmetrics.MetricRequestTotal("myapp"))
    prm.AddMetric(true, webmetrics.MetricRequestURITotal("myapp"))
    prm.AddMetric(true, webmetrics.MetricStatusCodeTotal("myapp"))
    
    // Setup Gin
    router := gin.Default()
    
    // Add Prometheus middleware (collects metrics)
    router.Use(gin.HandlerFunc(prm.MiddleWareGin))
    
    // Expose metrics endpoint
    router.GET("/metrics", prm.ExposeGin)
    
    // Your API routes
    router.GET("/api/users", handleUsers)
    router.POST("/api/orders", handleOrders)
    
    router.Run(":8080")
}

func handleUsers(c *gin.Context) {
    c.JSON(200, gin.H{"users": []string{"alice", "bob"}})
}

func handleOrders(c *gin.Context) {
    c.JSON(201, gin.H{"order_id": "12345"})
}

Complete Setup with All Metrics

package main

import (
    "github.com/gin-gonic/gin"
    libctx "github.com/nabbar/golib/context"
    "github.com/nabbar/golib/prometheus"
    "github.com/nabbar/golib/prometheus/webmetrics"
)

func setupMetrics(prm prometheus.Prometheus) {
    prefix := "myapp"
    getFct := func() prometheus.Prometheus { return prm }
    
    // Traffic metrics
    prm.AddMetric(true, webmetrics.MetricRequestTotal(prefix))
    prm.AddMetric(true, webmetrics.MetricRequestURITotal(prefix))
    prm.AddMetric(true, webmetrics.MetricRequestIPTotal(prefix))
    
    // Performance metrics
    prm.AddMetric(true, webmetrics.MetricRequestLatency(prefix, getFct))
    prm.AddMetric(true, webmetrics.MetricRequestSlow(prefix, getFct))
    
    // Error metrics
    prm.AddMetric(true, webmetrics.MetricRequestErrors(prefix))
    prm.AddMetric(true, webmetrics.MetricStatusCodeTotal(prefix))
    
    // Resource metrics
    prm.AddMetric(true, webmetrics.MetricRequestBody(prefix))
    prm.AddMetric(true, webmetrics.MetricResponseBody(prefix))
    prm.AddMetric(true, webmetrics.MetricResponseSizeByEndpoint(prefix, getFct))
    
    // Active connections (requires custom middleware)
    activeConns := webmetrics.MetricActiveConnections(prefix)
    prm.AddMetric(true, activeConns)
    
    // Configure request duration buckets
    prm.SetDuration([]float64{0.1, 0.3, 1.2, 5, 10})
}

func main() {
    prm := prometheus.New(libctx.NewContext)
    prm.SetSlowTime(5)
    
    setupMetrics(prm)
    
    // Exclude health check endpoints from metrics
    prm.ExcludePath("/health", "/ready", "/metrics")
    
    router := gin.Default()
    router.Use(gin.HandlerFunc(prm.MiddleWareGin))
    router.GET("/metrics", prm.ExposeGin)
    
    // Health endpoints (excluded from metrics)
    router.GET("/health", func(c *gin.Context) {
        c.JSON(200, gin.H{"status": "healthy"})
    })
    
    // API routes (monitored)
    api := router.Group("/api/v1")
    {
        api.GET("/users", handleUsers)
        api.POST("/users", createUser)
        api.GET("/users/:id", getUser)
    }
    
    router.Run(":8080")
}

Custom Metric Registration

package main

import (
    "context"
    libctx "github.com/nabbar/golib/context"
    "github.com/nabbar/golib/prometheus"
    prmmet "github.com/nabbar/golib/prometheus/metrics"
    prmtps "github.com/nabbar/golib/prometheus/types"
)

func main() {
    prm := prometheus.New(libctx.NewContext)
    
    // Create custom counter
    counter := prmmet.NewMetrics("custom_events_total", prmtps.Counter)
    counter.SetDesc("Total number of custom events")
    counter.AddLabel("event_type", "severity")
    counter.SetCollect(func(ctx context.Context, metric prmmet.Metric) {
        // Collection logic here
        metric.Inc(map[string]string{
            "event_type": "login",
            "severity": "info",
        })
    })
    
    // Create custom histogram
    histogram := prmmet.NewMetrics("task_duration_seconds", prmtps.Histogram)
    histogram.SetDesc("Task processing duration in seconds")
    histogram.AddLabel("task_type")
    histogram.AddBuckets(0.1, 0.5, 1.0, 2.5, 5.0, 10.0)
    histogram.SetCollect(func(ctx context.Context, metric prmmet.Metric) {
        // Collection logic here
        metric.Observe(map[string]string{"task_type": "import"}, 1.5)
    })
    
    // Register custom metrics
    prm.AddMetric(false, counter)    // false = not API metric
    prm.AddMetric(false, histogram)
    
    // Manually trigger collection
    prm.CollectMetrics(context.Background())
}

---

Subpackages

metrics Subpackage

Metric definitions with collection logic and Prometheus collector integration.

Features

• Type-safe metric creation (Counter, Gauge, Histogram, Summary)
• Label support for dimensional metrics
• Collection function callbacks
• Prometheus collector registration
• Helper methods (Inc, Dec, Set, Observe)

API Example

import prmmet "github.com/nabbar/golib/prometheus/metrics"
import prmtps "github.com/nabbar/golib/prometheus/types"

// Create counter
counter := prmmet.NewMetrics("http_requests_total", prmtps.Counter)
counter.SetDesc("Total HTTP requests")
counter.AddLabel("method", "status")
counter.SetCollect(func(ctx context.Context, metric prmmet.Metric) {
    // Increment counter
    metric.Inc(map[string]string{"method": "GET", "status": "200"})
})

// Create histogram
latency := prmmet.NewMetrics("http_duration_seconds", prmtps.Histogram)
latency.SetDesc("HTTP request latency")
latency.AddLabel("endpoint")
latency.AddBuckets(0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10)
latency.SetCollect(func(ctx context.Context, metric prmmet.Metric) {
    // Observe duration
    metric.Observe(map[string]string{"endpoint": "/api/users"}, 0.123)
})

Metric Types

• Counter: Monotonically increasing values (use Inc, Add)
• Gauge: Values that go up/down (use Inc, Dec, Set, Add, Sub)
• Histogram: Distributions with buckets (use Observe)
• Summary: Streaming quantiles (use Observe)

See GoDoc for complete API.

---

pool Subpackage

Thread-safe metric pool management with efficient storage and retrieval.

Features

• Add/Get/Delete operations
• List all registered metrics
• Walk through metrics with callbacks
• Thread-safe with sync.Map
• Context-aware operations

API Example

import prmpool "github.com/nabbar/golib/prometheus/pool"
import libctx "github.com/nabbar/golib/context"

// Create pool
pool := prmpool.New(libctx.NewContext)

// Add metric
err := pool.Add(metric)

// Retrieve metric
m := pool.Get("metric_name")

// List all metrics
names := pool.List()

// Walk through metrics
pool.Walk(func(p prmpool.MetricPool, key string, val prmmet.Metric) bool {
    fmt.Printf("Metric: %s\n", key)
    return true // continue walking
})

// Delete metric
pool.Del("metric_name")

Thread Safety: All operations are safe for concurrent use

See GoDoc for complete API.

---

types Subpackage

Core type definitions, interfaces, and Prometheus collector registration.

Features

• Metric interface definition
• MetricType enum (Counter, Gauge, Histogram, Summary)
• Prometheus collector creation
• Type validation
• GoDoc-documented examples

API Example

import prmtps "github.com/nabbar/golib/prometheus/types"

// Metric types
var (
    None      prmtps.MetricType = 0  // Invalid/uninitialized
    Counter   prmtps.MetricType = 1  // Cumulative
    Gauge     prmtps.MetricType = 2  // Up/down
    Histogram prmtps.MetricType = 3  // Distribution
    Summary   prmtps.MetricType = 4  // Quantiles
)

// Register metric with Prometheus
collector, err := metricType.Register(metric)
if err != nil {
    log.Fatal(err)
}
prometheus.MustRegister(collector)

Type-Specific Requirements

• Counter/Gauge: No special requirements
• Histogram: Must call AddBuckets() before registration
• Summary: Must call SetObjectives() before registration

See GoDoc for complete API.

---

webmetrics Subpackage

Pre-configured production-ready metrics for web servers.

11 Production Metrics

1. request_total - Total request count
2. request_uri_total - Requests per endpoint
3. request_duration - Request latency distribution
4. request_slow_total - Slow requests tracking
5. request_errors_total - Error tracking
6. request_body_total - Incoming bandwidth
7. response_body_total - Outgoing bandwidth
8. response_size_bytes - Response size distribution
9. request_ip_total - Unique client IPs
10. status_code_total - Status code distribution
11. active_connections - Current active requests

Quick Setup

import "github.com/nabbar/golib/prometheus/webmetrics"

prefix := "myapp"
getFct := func() prometheus.Prometheus { return prm }

// Essential metrics
prm.AddMetric(true, webmetrics.MetricRequestTotal(prefix))
prm.AddMetric(true, webmetrics.MetricRequestURITotal(prefix))
prm.AddMetric(true, webmetrics.MetricStatusCodeTotal(prefix))
prm.AddMetric(true, webmetrics.MetricRequestErrors(prefix))

// Performance metrics
prm.AddMetric(true, webmetrics.MetricRequestLatency(prefix, getFct))
prm.AddMetric(true, webmetrics.MetricRequestSlow(prefix, getFct))

// Bandwidth metrics
prm.AddMetric(true, webmetrics.MetricRequestBody(prefix))
prm.AddMetric(true, webmetrics.MetricResponseBody(prefix))
prm.AddMetric(true, webmetrics.MetricResponseSizeByEndpoint(prefix, getFct))

// Client tracking
prm.AddMetric(true, webmetrics.MetricRequestIPTotal(prefix))

// Active connections (requires special middleware)
prm.AddMetric(true, webmetrics.MetricActiveConnections(prefix))

See webmetrics/README.md for detailed documentation, PromQL queries, and dashboard examples.

---

bloom Subpackage

Space-efficient probabilistic data structure for unique value tracking.

Features

• Fast membership testing
• Configurable false positive rate
• Memory efficient (uses bit array)
• Thread-safe operations
• No false negatives

Use Case: Track unique IP addresses without storing all IPs

API Example

import "github.com/nabbar/golib/prometheus/bloom"

// Create bloom filter (1000 items, 1% false positive rate)
bf := bloom.New(1000, 0.01)

// Add items
bf.Add([]byte("192.168.1.1"))
bf.Add([]byte("10.0.0.1"))

// Check membership
if bf.Test([]byte("192.168.1.1")) {
    // Probably present (might be false positive)
}

// Get approximate count
count := bf.Count()

Performance: O(k) where k is number of hash functions (~3-7 for 1% FPR)

See GoDoc for complete API.

---

Best Practices

Choose Appropriate Metric Types

// ✅ Good: Counter for cumulative values
requests := prmmet.NewMetrics("requests_total", prmtps.Counter)

// ✅ Good: Gauge for current state
activeConns := prmmet.NewMetrics("active_connections", prmtps.Gauge)

// ✅ Good: Histogram for distributions
latency := prmmet.NewMetrics("request_duration_seconds", prmtps.Histogram)
latency.AddBuckets(0.1, 0.3, 1.2, 5, 10)

// ❌ Bad: Gauge for cumulative values (will reset)
requests := prmmet.NewMetrics("requests_total", prmtps.Gauge)

Manage Label Cardinality

// ✅ Good: Low cardinality (bounded label values)
counter.AddLabel("method", "status", "endpoint")  // ~5 × ~50 × ~100 = 25K combinations

// ❌ Bad: High cardinality (unbounded label values)
counter.AddLabel("user_id", "timestamp")  // Millions of combinations!

Use Path Patterns, Not Actual Paths

// ✅ Good: Use route pattern
endpoint := c.FullPath()  // "/users/:id"

// ❌ Bad: Use actual path
endpoint := c.Request.URL.Path  // "/users/12345" (creates thousands of series)

Exclude Health Checks

// Prevent metrics pollution from health checks
prm.ExcludePath("/health", "/ready", "/metrics", "/favicon.ico")

Set Appropriate Histogram Buckets

// ✅ Good: Buckets appropriate for your use case
// For API latency (milliseconds to seconds)
latency.AddBuckets(0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10)

// For file sizes (bytes to megabytes)
size.AddBuckets(1024, 10240, 102400, 1048576, 10485760)

// ❌ Bad: Too many buckets (memory waste)
latency.AddBuckets(0.001, 0.002, 0.003, ..., 100)  // 100+ buckets!

Always Handle Errors

// ✅ Good: Check errors
if err := prm.AddMetric(true, metric); err != nil {
    log.Printf("Failed to add metric: %v", err)
}

// ❌ Bad: Ignore errors
prm.AddMetric(true, metric)

Clean Up Metrics

// Remove unused metrics to free memory
defer prm.DelMetric("temporary_metric")

---

Testing

Test Suite: 426 specs across 5 packages with 91.6% overall coverage

# Run all tests
go test ./...

# With coverage
go test -cover ./...

# With race detection (recommended)
CGO_ENABLED=1 go test -race ./...

Test Results

prometheus/                137 specs    90.9% coverage   1.38s
prometheus/bloom/           24 specs    94.7% coverage   0.13s
prometheus/metrics/        179 specs    95.5% coverage   0.03s
prometheus/pool/            74 specs    72.5% coverage   0.02s
prometheus/types/           36 specs   100.0% coverage   0.01s

Quality Assurance

• ✅ Zero data races (verified with -race)
• ✅ Thread-safe concurrent operations
• ✅ Comprehensive edge case testing
• ✅ Integration tests with Gin

See TESTING.md for detailed testing documentation.

---

Contributing

Contributions are welcome! Please follow these guidelines:

Code Contributions

• Do not use AI to generate package implementation code
• AI may assist with tests, documentation, and bug fixing
• All contributions must pass go test -race
• Maintain or improve test coverage (≥80%)
• Follow existing code style and patterns

Documentation

• Update README.md for new features
• Add GoDoc comments for all public APIs
• Provide examples for common use cases
• Keep TESTING.md synchronized with test changes

Testing

• Write tests for all new features
• Test edge cases and error conditions
• Verify thread safety with race detector
• Add comments explaining complex test scenarios

Pull Requests

• Provide clear description of changes
• Reference related issues
• Include test results and coverage
• Update documentation

See CONTRIBUTING.md for detailed guidelines.

---

Future Enhancements

Potential improvements for future versions:

Metric Features

• Custom aggregation functions
• Metric templating and cloning
• Metric groups for bulk operations
• Metric versioning for breaking changes
• Automatic metric expiration

Performance

• Metric caching layer
• Batch metric collection
• Async metric updates
• Memory-optimized storage
• Custom metric serialization

Integration

• OpenTelemetry bridge
• StatsD protocol support
• InfluxDB line protocol
• CloudWatch metrics export
• Datadog integration

Observability

• Built-in tracing integration
• Metric correlation with logs
• Anomaly detection
• Automatic alerting rules
• Performance profiling

Developer Experience

• Metric discovery API
• Real-time metric preview
• Metric validation CLI
• Dashboard templates
• Grafana integration

Suggestions and contributions are welcome via GitHub issues.

---

AI Transparency Notice

In accordance with Article 50.4 of the EU AI Act, AI assistance has been used for testing, documentation, and bug fixing under human supervision.

---

License

MIT License - See LICENSE file for details.

---

Resources

• Issues: GitHub Issues
• Documentation: GoDoc
• Testing Guide: TESTING.md
• Web Metrics: webmetrics/README.md
• Contributing: CONTRIBUTING.md
• Prometheus: prometheus.io
• Gin Framework: gin-gonic.com