Start and stop deployments

Start a deployment#

Creating a deploy starts it. When you choose Create deploy in Deployments→New deploy, DevLyft places the deploy in its location and starts the number of replicas you asked for.

To check it came up, look at the deploy's pods on the Pods page, where each replica shows a Phase, and open Watch logs on the deploy. See Deployment status.

Stop a deployment#

The dashboard has no Stop button, and Replicas can't be set below 1. To stop a workload today, delete its deploy:

1. Note its settings

   Open Edit on the deploy and note its build, location, replicas, tenancy and variables, so you can recreate it later. Open Routes and note any hostnames.

2. Delete it

   Choose ⋮ → Delete on the row and confirm.

Deleting a deploy keeps the pieces you'd need to bring it back:

• The build stays on the Builds page.
• Secrets stay in the environment.
• Verified domains stay on the Domains page. A route's CNAME target is derived from its hostname alone, so when you attach the same hostname again the target is unchanged and your DNS doesn't need touching.

Start it again#

Create a new deploy with the same build and settings from Deployments→New deploy, then attach its routes again.

Billing when stopped#

• Shared compute: billing is per second on running replicas, so when a workload stops, billing for it stops immediately. Bring it back up and billing resumes.
• Isolated compute: you pay for each dedicated node while it's provisioned, regardless of how much runs on it. Billing for a node stops once everything on it has been deprovisioned.

See Usage and billing.

Scale to zero#

Shared compute is described as scaling to zero when idle.

Restarts#

A replica that exits or runs out of memory is restarted, and each restart is counted. View metrics shows a Restarts series, and the dashboard flags a deploy once its restarts reach 3. It doesn't claim to know why a replica restarted, so check Watch logs for the reason.