Task Metrics using Prometheus in Go

Table of Contents

• Introduction
• Features
• Default Labels
• Code Structure
• Metrics Collectors
  • Counter
  • Gauge
  • Histogram
• Usage
  • Starting the Server
  • Registering Metrics
  • Updating Metrics

Introduction

This project provides a comprehensive metrics collection and reporting framework integrated with Prometheus for Go applications.

Features

• Different types of metrics collectors including Counter, Gauge, and Histogram.
• Built-in Prometheus server.
• Strong typing to prevent metrics misuse.
• Error handling and logging integrated with Uber's Zap library.
• Support for custom labels.
• Thread-safe metric updates with read-write mutexes.
• Auto-registering of metrics based on type.

Default Labels

Every metric reported by the framework automatically includes the following labels to provide more context about the application's environment:

• app_name: Identifies the name of the application.
• app_revision: Specifies the current revision of the application, typically a Git commit hash or a similar identifier.
• app_version: Indicates the current version of the application.

These labels help in distinguishing metrics across different environments, versions, and revisions of the application. For example, a metric for request duration might look like this:

request_duration_ms{app_name="ledger-live",app_revision="main-9a60456",app_version="v0.8.3",method="GET",path="/addresses/{address}/transactions",status="200"} 435

This approach ensures that metrics data can be correlated with specific releases and deployments of the application, providing valuable insights during analysis and troubleshooting.

Code Structure

• /metrics/collectors/: Contains the implementation of various metrics collectors (Counter, Gauge, Histogram).
• /metrics/handler.go: Contains the MetricHandler interface and taskMetrics UpdateMetric, IncrementMetric, and DecrementMetric methods for updating metrics.
• /metrics/prometheus.go: Defines the Prometheus server.
• /metrics/register.go: Responsible for registering metrics with the Prometheus server.

Metrics Collectors

Counter

• File: /metrics/collectors/counter.go
• Methods: Update
• Usage: Counters are cumulative metrics that can only increase.

Gauge

• File: /metrics/collectors/gauge.go
• Methods: Update, Inc, Dec
• Usage: Gauges are metrics that can arbitrarily go up and down.

Histogram

• File: /metrics/collectors/histogram.go
• Methods: Update
• Usage: Histograms count observations (like request durations or response sizes) and place them in configurable buckets.

Usage

Starting the Server

metricsServer := metrics.NewTaskMetrics("/metrics", "9090")
err := metricsServer.Start()
if err != nil {
    log.Fatal(err)
}

Registering Metrics

// Without labels
err = metricsServer.RegisterMetric("my_counter", "This is a counter metric", nil, &collectors.Counter{})

// With labels
err = metricsServer.RegisterMetric("my_counter_with_labels", "This is a counter metric with labels", []string{"label1", "label2"}, &collectors.Counter{})

// Register a gauge
err = metricsServer.RegisterMetric("my_gauge", "This is a gauge metric", nil, &collectors.Gauge{})

Updating Metrics

// Without labels
metricsServer.UpdateMetric("my_counter", 1)

// With labels
metricsServer.UpdateMetric("my_counter_with_labels", 1, "label1_value", "label2_value")

// Increment a gauge
metricsServer.IncrementMetric("my_gauge")

// Decrement a gauge
metricsServer.DecrementMetric("my_gauge")

Mocking Support

Use MockTaskMetrics for testing.

func TestMetrics(t *testing.T) {
    tm := &metrics.MockTaskMetrics{}
	  tm.On("RegisterMetric", "local_cache_cleanup_errors", mock.Anything, []string{"error_type"}, mock.Anything).Once().Return(nil)
    // Use tm in your tests
}

To generate mocks:

• install: https://github.com/vektra/mockery
• pull the repo with the correct version of the interface you want to mock
• mockery --name TaskMetrics --dir ./pkg/metrics --output . --filename task_metrics_mock.go --structname MockTaskMetrics --inpackage
• For usage in tests