CloudSQL Autoscaler

Website: https://fraser-isbester.github.io/cloudsql-autoscaler/

Automatically scales Google Cloud SQL instances based on CPU and memory utilization patterns. Supports both one-shot CLI usage and continuous daemon mode for production deployments.

Features

• Smart Analysis: Uses 3 days of metrics and P95 percentiles to avoid scaling on temporary spikes
• Enterprise Aware: Understands downtime constraints for Enterprise vs Enterprise Plus editions
• Multiple Modes: CLI tool, Docker container, or Kubernetes deployment
• Configurable: Conservative, default, and aggressive scaling profiles
• Observable: Built-in Prometheus metrics and health endpoints

Quick Start

# Install
go install github.com/fraser-isbester/cloudsql-autoscaler/cmd/cloudsql-autoscaler@latest

# Analyze all instances (dry-run by default)
cloudsql-autoscaler --project my-gcp-project

# Apply scaling recommendations
cloudsql-autoscaler --project my-gcp-project --dry-run=false

# Analyze specific instances
cloudsql-autoscaler --project my-gcp-project --instance db1 --instance db2

Usage Options

Basic Commands

# Core flags
--project string       GCP project ID
--instance strings     Specific instance(s) to analyze (default: all)
--dry-run             Show recommendations without applying (default: true)
--output string       Format: table or json (default: table)

# Daemon mode for continuous operation
--daemon              # Run continuously
--interval duration   # Check interval (default: 30m)
--http-port int       # Health/metrics port (default: 8080)

Example Commands

# One-shot analysis with JSON output
cloudsql-autoscaler --project my-project --output json

# Conservative scaling for production
cloudsql-autoscaler --project my-project --profile conservative --dry-run=false

# Continuous monitoring with 15-minute intervals
cloudsql-autoscaler --daemon --project my-project --interval=15m

Deployment Options

Docker

# Pull and run
docker pull ghcr.io/fraser-isbester/cloudsql-autoscaler:latest
docker run -d \
  -e GOOGLE_APPLICATION_CREDENTIALS=/creds.json \
  -v /path/to/creds.json:/creds.json:ro \
  ghcr.io/fraser-isbester/cloudsql-autoscaler:latest \
  --daemon --project my-gcp-project

Kubernetes

# Clone and deploy
git clone https://github.com/fraser-isbester/cloudsql-autoscaler.git
cd cloudsql-autoscaler
make deploy-k8s

# Check status
make k8s-status
make k8s-logs

Prerequisites for Kubernetes:

• GKE cluster with Workload Identity enabled
• Google Cloud Service Account with Cloud SQL Admin permissions

See deploy/kubernetes/README.md for complete setup instructions.

Monitoring

When running in daemon mode, health and metrics endpoints are available:

curl http://localhost:8080/health   # Health check
curl http://localhost:8080/ready    # Readiness probe
curl http://localhost:8080/status   # Detailed status
curl http://localhost:8080/metrics  # Prometheus metrics

Key Metrics:

• cloudsql_autoscaler_instances_total - Total instances in project
• cloudsql_autoscaler_instances_scalable - Instances needing scaling
• cloudsql_autoscaler_scaling_operations_total - Scaling operations by result
• cloudsql_autoscaler_cycle_duration_seconds - Analysis cycle duration

How it Works

1. Collects Metrics: Gathers 3 days of CPU/memory data from Cloud Monitoring
2. Analyzes Patterns: Uses P95 percentiles to identify sustained load vs spikes
3. Recommends Changes: Suggests machine type upgrades/downgrades within constraints
4. Respects Limits: Understands Enterprise Plus zero-downtime windows vs Enterprise downtime requirements
5. Applies Safely: Optionally executes changes with proper error handling and rollback

Supported Machine Types:

• Standard: db-f1-micro, db-g1-small, db-n1-*, db-n2-*, db-e2-*
• Custom: db-custom-{vcpus}-{memory_mb}
• Performance: db-perf-optimized-N-*