monitoring

Documentation

Overview

Package monitoring publishes CortexOps custom metrics to AWS CloudWatch.

Metrics are published under the "CortexOps" namespace after each observe → reason → record cycle. When running locally without AWS credentials, all publish calls silently no-op.

Index

• type CycleMetrics
• type Monitor
  • func NewMonitor(ctx context.Context) *Monitor
  • func (m *Monitor) PublishCycleMetrics(ctx context.Context, metrics CycleMetrics)

Types

type CycleMetrics

type CycleMetrics struct {
	CycleSuccess  bool    // true if cycle completed and recorded a decision
	CPUPercent    float64 // cluster CPU % from MCP snapshot
	ActiveQueries int     // active query count from MCP snapshot
	Confidence    float64 // AI decision confidence (0.0–1.0)
	ActionType    string  // e.g. "backup", "no_action", "scale_up"
}

CycleMetrics holds the telemetry data collected from one observe cycle.

type Monitor

type Monitor struct {
	// contains filtered or unexported fields
}

Monitor wraps the CloudWatch client for metric publishing.

func NewMonitor

func NewMonitor(ctx context.Context) *Monitor

NewMonitor creates a CloudWatch monitor. If AWS credentials are not available (e.g. local dev), it returns a no-op monitor that logs metrics locally instead of publishing to CloudWatch.

func (*Monitor) PublishCycleMetrics

func (m *Monitor) PublishCycleMetrics(ctx context.Context, metrics CycleMetrics)

PublishCycleMetrics sends custom metrics to CloudWatch after each cycle. If CloudWatch is unavailable, metrics are logged to stdout instead.