Fly.io

Overview

Fly.io deploys applications to Firecracker microVMs across 30+ edge regions worldwide, providing sub-50ms latency to users. It supports scale-to-zero machines, persistent NVMe volumes, LiteFS for multi-region SQLite replication, and private WireGuard networking between services.

Instructions

• When deploying applications, use fly launch to auto-detect the framework and generate a Dockerfile, then fly deploy for zero-downtime rolling updates with health checks.
• When configuring scaling, use auto_stop_machines and auto_start_machines in fly.toml to scale to zero when idle and wake on incoming requests, and set machine sizing appropriate to the workload.
• When managing multi-region deployments, use fly scale count --region to distribute machines, fly-replay header to route writes to the primary region, and LiteFS for SQLite read replicas.
• When handling persistent data, attach volumes for durable storage (machines are ephemeral), use LiteFS for multi-region SQLite, or Tigris for S3-compatible object storage.
• When connecting services, use .internal DNS for private service-to-service communication over the WireGuard mesh and never expose internal services to the public internet.
• When managing secrets, use fly secrets set KEY=value for encrypted secret storage accessible as environment variables.
• When troubleshooting, use fly logs for real-time streaming, fly ssh console to access running machines, and fly proxy to tunnel to internal services.

Examples

Example 1: Deploy a multi-region web application