Deploy

Zero-downtime deploys

Roll out new versions behind health checks and roll back in one click.

Start for free Talk to sales

Deploy logs

10:12:00build 1042 started
10:13:01image pushed in 21s
10:14:023 new instances up
10:15:03health 200 on 3 of 3
10:16:04traffic moved to 1042
10:17:05build 1041 drained
10:18:06build 1042 is live
10:19:07GET /api/orders 200
10:12:08build 1042 started
10:13:09image pushed in 21s
10:14:003 new instances up
10:15:01health 200 on 3 of 3
10:16:02traffic moved to 1042
10:17:03build 1041 drained
10:18:04build 1042 is live
10:19:05GET /api/orders 200

Requests

Ship on a Friday afternoon

New versions only receive traffic once they are healthy, and old versions stay ready until the switch is done.

Health-checked rollouts
Traffic moves only after the new instances pass your health check.
Graceful draining
In-flight requests finish on the old version before it shuts down.
One-click rollback
Return to any previous build instantly, without rebuilding.
Pre-deploy commands
Run migrations or cache warmups before any traffic shifts.

How a deploy rolls out

Every push follows the same safe sequence, from build to live.

Git push
Build
Migrate
Release

Configuration

Tune the rollout

Set a health check path, a pre-deploy command, and how long old instances should drain.

Custom health check path and timeout
Migrations before traffic shifts
Rollback from the CLI or dashboard

          
            
                1
                services:
              
                2
                  - type: web
              
                3
                    name: storefront
              
                4
                    healthCheckPath: /healthz
              
                5
                    preDeployCommand: npm run db:migrate
              
                6
                    deploy:
              
                7
                      drainSeconds: 30
              
                8
                      maxSurgePercent: 100

          
            
                1
                # Roll back to the previous healthy build
              
                2
                $ cloud rollback storefront --to 1041
              
                3
                → Routing traffic to build 1041
              
                4
                ✓ storefront is live on build 1041

Why deploys break, and how we avoid it

Most outages during a release come from the gap between stopping the old version and starting the new one, or from a new version that starts but cannot serve requests. Zero-downtime deploys close that gap: the old version keeps serving until the new one proves it is healthy.

What happens on every deploy

You do not need to configure anything to get safe rollouts. A default health check runs against every web service, and you can point it at your own endpoint for a stricter check.

New instances start before old ones stop
Traffic shifts only after health checks pass
Failed deploys never receive traffic

Rolling back

Every successful build is kept as a rollback target. Rolling back routes traffic to an existing image, so it takes seconds instead of a full rebuild. Database migrations are not reversed automatically, so keep them backward compatible.

Zero-downtime deploys

Ship on a Friday afternoon

Health-checked rollouts

Graceful draining

One-click rollback

Pre-deploy commands

How a deploy rolls out

Tune the rollout

Why deploys break, and how we avoid it

What happens on every deploy

Rolling back

Works well with

Preview environments

Autoscaling

CLI and API

Push code today. Be live before your coffee cools.