Skip to main content
A deploy (deployment) is a single release of application code to one module instance. Where the stack tracks a module’s infrastructure changes, deploys track its code releases. The two are separate: a deploy never runs Terraform, and a stack change never redeploys your code. Each deploy records:
  • A description and the artifact inputs it was given, such as a digest-pinned image URI.
  • A status: PENDING → IN_PROGRESS → SUCCESS, ERROR, or CANCELLED (plus ROLLING_BACK and ROLLED_BACK).
  • A timeline of deployment events and instance-level resource status.
  • A snapshot of what was released — for ECS, the full task definition — which is what rollback replays.

​
Deployment types

The module’s definition determines how a deploy rolls out: ECS deploys also support pre_deploy and post_deploy hooks — one-off tasks like database migrations that run before or after the rollout. A failed pre-deploy hook aborts the deploy. EC2 deploys roll out across the group with configurable concurrency and an error threshold that stops the rollout early. Instances launched later by autoscaling or instance replacement automatically run the active release so they catch up. See rvn-ec2-service. Deploys are locked per module instance, so concurrent triggers queue instead of colliding.

​
How deploys are triggered

1

From a Ravion pipeline

A deploy step passes a preceding build step’s output to the module’s deploy manager:
The deploy records which pipeline run and step created it. See step types.
2

From the API or CLI

External CI systems like GitHub Actions build the artifact themselves, then trigger the deploy:
3

From the dashboard

Trigger a deploy manually from the module instance page.

​
Watching a deploy

Ravion surfaces every low-level event during a rollout — far more than a CI log line:
  • Deploy phases: provision, snapshot, pre-deploy hook, update, wait-for-healthy, post-deploy hook, activate, complete — with traffic-shift, bake, and pause sub-phases for blue-green style strategies.
  • Resource-level status: each ECS task’s status ladder (provisioning → pending → running) and target-group health.
  • AWS infrastructure events: task state changes, service actions, and deployment state changes ingested from EventBridge.
All of this streams live to the dashboard and is available via the API and ravion deploy get.

​
Rollback, redeploy, cancel

  • Rollback starts a new deploy that replays a previous successful deploy’s snapshot. List valid targets (previous successes whose images still exist in the registry), then roll back:
  • Redeploy (ravion deploy redeploy <id>) starts a new deploy with the same definition as a previous one.
  • Cancel (ravion deploy cancel <id>) stops an in-progress deploy; ECS traffic-shift deploys still in their rollback window are rolled back natively by ECS.
  • Continue (ravion deploy continue <id>) resumes a deploy paused on a pause-stage hook.
See the ravion deploy CLI reference.