Roll back a deploy

Roll back by hand

Every deploy pins an explicit digest, so a rollback is a deploy of the earlier digest.

1. Find the digest you want. statio logs api lists one line per deploy, each with its digest.
2. Open the repo’s Actions tab, pick the deploy workflow, and choose Run workflow.
3. Paste the digest into the digest input and run it.

CI signs a fresh payload carrying the old digest. The agent treats it as a new deploy, so the sequence check passes.

The generated deploy.yml ships this input. If you wrote your own workflow, add it:

on:
  workflow_dispatch:
    inputs:
      digest:
        description: "Existing image digest to redeploy"
        required: false

and pass it through:

with:
  digest: ${{ inputs.digest }}

What the agent does on its own

When a new deploy fails its health check, the agent restores the previous image and its env together, as one unit, and answers failure_rolled_back. The snapshot lives in /run, on tmpfs, so it survives until the server reboots.

After a reboot the agent has no snapshot to restore. Redeploy from CI.

Why nothing broke publicly

The agent probes health before it touches the proxy or DNS, so a broken deploy never reaches the edge, and a rollback leaves the edge untouched.