Skip to content

Rollbacks

Falak keeps your recent releases on every server, so going back is a switch, not a rebuild. A rollback is a deployment with trigger rollback that activates an earlier release.

  1. Open the site’s Deployments tab.
  2. On the deployment you want to return to, open ⋯ → Rollback. Or use ⋯ → Rollback… in the panel header to pick a release.
  3. Confirm. The rollback streams like any deployment.

The Releases list of a site: retained releases with commit, branch, author and activation time; the active release is marked and older ones offer a rollback action.

A deployment rolls back by itself when a step fails after some servers switched — for example a failed health check. The deployment ends failed with rolled_back: true, the deployments.rolled_back alert fires, and every switched server serves the previous release again.

If a step fails before activation (build, fetch, prepare, migrate), nothing switched, so there is nothing to roll back.

Restored Not restored
The release’s code Database migrations (Falak never runs migrate:rollback)
The environment variables the release was deployed with Files in shared paths (storage/, uploads)
Processes restarted with that release’s environment External side effects of the newer release
Docker: the previous image; Compose: the previous compose.yaml, .env and pinned digests Compose named volumes
  • Releases are kept per site, 5 by default (Settings → Deploy → Releases to keep, 1–50).
  • GET /api/v1/sites/{site}/releases lists them, current first, with can_rollback.
  • A failed or pruned release cannot be activated.