Skip to content

README.md

Latest commit

 

History

History
164 lines (116 loc) · 3.55 KB

README.md

File metadata and controls

164 lines (116 loc) · 3.55 KB

1000 — Zero-Downtime Deploys

How Fly Deploys Work

When you run fly deploy, Fly performs a rolling deployment by default:

Old Machine v1 (running)
New Machine v2 (starting)
    │
    ▼ health check passes
New Machine v2 (healthy)
    │
    ▼ Fly proxy shifts traffic
Old Machine v1 (stopped / destroyed)

At no point is there a gap in service — the old Machine handles traffic until the new one is proven healthy.

Deployment Strategies

Rolling (default)

Replaces Machines one at a time. Zero downtime.

fly deploy

Bluegreen

Creates all new Machines before cutting over. More resource-intensive but instant cutover.

fly deploy --strategy bluegreen

Canary

Sends a small % of traffic to the new version before full rollout.

fly deploy --strategy canary

Immediate

Stops old, starts new without health check waiting (fastest but brief downtime).

fly deploy --strategy immediate

Health Checks & Deploy Safety

Fly won't mark a Machine as healthy (and won't cut over traffic) until health checks pass.

[[http_service.checks]]
  grace_period = "5s"     # Wait before checking after start
  interval = "10s"
  method = "GET"
  path = "/health"
  timeout = "2s"

Your app must return HTTP 200 on /health before Fly considers the deploy successful.

Release Commands

Run a command before new Machines start (e.g., database migrations):

[deploy]
  release_command = "python manage.py migrate"

The release command runs in a temporary Machine using the new image. If it fails (non-zero exit), the deploy is aborted.

This is the correct pattern for database migrations — run migrations before routing traffic to new app code.

Rollback

Instant rollback to previous image

fly deploy --image registry.fly.io/my-app:<previous-deployment-tag>

List previous deployments

fly releases

Output:

VERSION  STABLE  TYPE     STATUS   DESCRIPTION           USER     DATE
v10      true    release  complete deploy                 user@ex  1h ago
v9       true    release  complete deploy                 user@ex  3h ago

Rollback to a specific version

fly deploy --image registry.fly.io/my-app:deployment-01ABCD

Wait for Deploy

fly deploy --wait-timeout 120   # Wait up to 120s for deploy to complete

Monitoring a Deploy

fly logs           # Watch live logs during deploy
fly status         # Check Machine states

Deploy from a Specific Docker Image

fly deploy --image nginx:1.25
fly deploy --image ghcr.io/myorg/myapp:v1.2.3

Tasks

Task 1 — Observe a rolling deploy

  1. Deploy your app.
  2. Modify the app (change a response message).
  3. Run fly deploy and in a parallel terminal run fly logs — observe the deploy process.
  4. In a third terminal, curl your app repeatedly during the deploy: watch -n 0.5 curl -s https://my-app.fly.dev — confirm no errors.

Task 2 — Add a release command

  1. Add a simple release command (e.g., echo "Migration complete") to fly.toml: 
    [deploy]
  release_command = "echo 'Running migrations...'"
  2. Deploy and watch the logs — you should see the release command output.

Task 3 — Rollback

  1. Make a breaking change to your app.
  2. Deploy it.
  3. Run fly releases to find the previous version.
  4. Roll back with fly deploy --image registry.fly.io/<app>:<old-tag>.

Next Steps

→ Continue to 1100 — Volumes & Persistent Storage