Fly.io Scale-to-Zero Configuration

Overview

Scale-to-zero is a Fly.io feature that automatically shuts down VMs after a period of inactivity to save costs in non-production environments. VMs automatically restart when new traffic arrives.

Activity is measured as: Traffic via the Fly proxy (flycast)

Key behaviors:

VMs shut down after configured inactivity period
VMs auto-restart when requests arrive
First request after wake-up will be slower (cold start)
Ongoing requests after wake-up should perform normally

Configuration

Scale-to-zero is configured differently for generic applications vs. Postgres databases.

Generic Applications

Add these settings to your fly.toml:

[http_service]
auto_stop_machines = true
auto_start_machines = true
min_machines_running = 0

What this means:

auto_stop_machines = true - Fly will automatically stop idle machines
auto_start_machines = true - Fly will automatically start stopped machines when traffic arrives
min_machines_running = 0 - Allow all machines to stop when idle

Postgres Applications

Add these settings to your Postgres app fly.toml:

[env]
FLY_SCALE_TO_ZERO = '1h'

[[services]]
protocol = 'tcp'
internal_port = 5432
auto_stop_machines = true
auto_start_machines = true

What this means:

FLY_SCALE_TO_ZERO = '1h' - Shut down after 1 hour of inactivity
Service configuration includes the same auto-start/stop settings as generic apps

Common Issues

Wake-from-Zero Reliability

Scale-to-zero wake-up doesn’t always work reliably. If you experience issues:

Check machine status: fly status --all -a your-app-name
Manually restart if needed: fly apps restart your-app-name

Inconsistent Shutdown Timing

Postgres instances may shut down faster than the configured FLY_SCALE_TO_ZERO timeout. This appears to be a Fly.io platform issue.

Disabling Scale-to-Zero

To disable scale-to-zero for a specific machine:

fly machines update $machine_id --env FLY_SCALE_TO_ZERO=""

When to Use Scale-to-Zero

Good for:

Staging environments
Demo environments
Development environments
Any environment with intermittent usage

Not recommended for:

Production environments
Applications requiring consistent response times
High-availability scenarios

Troubleshooting Commands

# Check app health
fly checks list -a your-app-name

# View all machines and their status
fly status --all -a your-app-name

# List machines
fly machines list -a your-app-name

# Manually start a specific machine
fly machines start machine_id