Skip to content

fly-cheatsheet.md

Latest commit

 

History

History
192 lines (134 loc) · 5.41 KB

fly-cheatsheet.md

File metadata and controls

192 lines (134 loc) · 5.41 KB

Fly.io Cheat Sheet

Quick reference for common operations on the Fly.io environments.

Setup

# Authenticate (token is in .env — uncomment it, or export inline)
export FLY_API_TOKEN="<token from .env>"

# Or log in interactively
flyctl auth login

Run an Artisan command on a live instance

flyctl ssh console --app restarters-dev --command "php /var/www/artisan <command>"

Examples:

# Add a network coordinator
flyctl ssh console --app restarters-dev \
  --command "php /var/www/artisan network:coordinator:add 'some@email.com'"

# Run migrations
flyctl ssh console --app restarters-dev \
  --command "php /var/www/artisan migrate --force"

# Check migration status
flyctl ssh console --app restarters-dev \
  --command "php /var/www/artisan migrate:status"

# Clear caches
flyctl ssh console --app restarters-dev \
  --command "php /var/www/artisan cache:clear"

# Open a PHP REPL (tinker)
flyctl ssh console --app restarters-dev --command "php /var/www/artisan tinker"

For production (restarters-prod), replace restarters-dev with restarters.

Open an interactive shell

flyctl ssh console --app restarters-dev

Deploy

# Deploy dev (normally triggered by CI on develop branch)
flyctl deploy --config fly.dev.toml --remote-only

# Deploy production
flyctl deploy --config fly.toml --remote-only

Logs

# Live log stream
flyctl logs --app restarters-dev

# Nginx access log (persistent — survives redeploys)
flyctl ssh console --app restarters-dev --command "tail -f /var/log/nginx/access.log"

# Laravel log (persistent — survives redeploys)
flyctl ssh console --app restarters-dev --command "tail -f /var/log/laravel/laravel.log"

# PHP-FPM slow log
flyctl ssh console --app restarters-dev --command "tail -f /var/log/php-fpm-slow.log"

Log format

The nginx access log uses a custom timed format (the X-Forwarded-For real client IP is logged first, then the Fly internal proxy IP in parentheses):

<real-client-ip> (<fly-proxy-ip>) - <user> [timestamp] "REQUEST" status bytes "referer" "ua"
rt=<total_request_time> uct=<fpm_connect_time> uht=<time_to_first_header> urt=<fpm_response_time>

Key fields:

  • rt = total time nginx held the request (includes FPM wait + processing + response send)
  • urt = time PHP-FPM took to process the request (the PHP execution time)
  • uct = time to connect to FPM socket (near 0 normally; high values indicate socket backlog)
  • uht = time until FPM sent the first response header (FPM queue wait + PHP start time)
  • uht=- = FPM never responded (504 / worker died)

Find slow requests (> 2s upstream time)

flyctl ssh console --app restarters --command \
  'awk '"'"'{for(i=1;i<=NF;i++) if($i~/^urt=/) {v=substr($i,5); if(v+0>2) print v,$0}}'"'"' \
  /var/log/nginx/access.log | sort -rn | head -30'

Identify queued requests (rt much larger than urt)

Queueing shows as uct (FPM connect time) being elevated. When FPM workers are all busy, the socket accept is delayed, so uct >> 0. The window uht - uct is the PHP processing time.

# Show requests where total time is > 5s — look for uct > 0 to spot FPM saturation
flyctl ssh console --app restarters --command \
  'awk '"'"'{for(i=1;i<=NF;i++) if($i~/^rt=/) {v=substr($i,3); if(v+0>5) print v,$0}}'"'"' \
  /var/log/nginx/access.log | sort -rn | head -30'

Summarise slowest URLs by average upstream time

flyctl ssh console --app restarters --command \
  'awk '"'"'{url="-"; urt=0; for(i=1;i<=NF;i++){if($i~/^"(GET|POST|PUT|DELETE|PATCH|HEAD)/) url=$i; if($i~/^urt=/) urt=substr($i,5)+0} if(urt>0) print urt, url}'"'"' \
  /var/log/nginx/access.log | sort -rn | head -40'

Top calling hosts (real IP)

The first field in the timed log is the real client IP (from X-Forwarded-For). The Fly proxy IP is field 2 (inside parens, stripped on parsing).

flyctl ssh console --app restarters --command \
  'awk '"'"'{print $1}'"'"' /var/log/nginx/access.log | sort | uniq -c | sort -rn | head -20'

Count 5xx errors by URL

# Fields: 1=real-ip 2=(fly-ip) 3=- 4=- 5=[timestamp] 6=timezone] 7="METHOD 8=PATH 9=PROTO" 10=status
flyctl ssh console --app restarters --command \
  'awk '"'"'$10~/^5/ {print $10, $8}'"'"' /var/log/nginx/access.log | sort | uniq -c | sort -rn | head -20'

Secrets (environment variables)

# List secrets (names only — values never shown)
flyctl secrets list --app restarters-dev

# Set a secret
flyctl secrets set MY_VAR="value" --app restarters-dev

# Delete a secret
flyctl secrets unset MY_VAR --app restarters-dev

Scale / machines

# List machines and their status
flyctl machine list --app restarters-dev

# Check app status
flyctl status --app restarters-dev

Database access (phpMyAdmin)

# Proxy to phpMyAdmin (opens at http://localhost:8080)
flyctl proxy 8080:80 --app restarters-pma

File system notes

  • The container filesystem is ephemeral — files written outside /var/log are lost on redeploy.
  • /var/log is a persistent volume (app_logs_dev) — logs and file cache survive redeploys.
  • Uploaded images are written to /var/www/public/uploads/ temporarily, then synced to Tigris S3.
    Nginx proxies /uploads/* → Tigris, so images are served from S3, not local disk.
  • public/uploads/ is created at build time (Dockerfile.fly) and on each container start (startup.sh).