Turtini

How Change Control decides to continue, stop or roll back

Change Control runs your own Ansible Automation Platform job template a wave of servers at a time. After each wave it decides one of three things.

**Continue.** The wave's job finished, the bake window passed, and every health check read every server in the wave and they all passed. The next wave starts, or waits for an approver if the plan asks for approval on every wave.

**Roll back.** The job failed on servers in the wave, or the wave fails a check that the servers outside the wave pass. Your rollback template runs on every server the change has reached, and no later wave starts.

**Stop for a person.** Any of these holds the change:
- a check returned nothing for a server in the wave (a timeout, Datadog unreachable, a server that did not report);
- the wave fails a check and the servers outside the wave fail it too, so the environment is unhealthy and the change is not blamed;
- the controller could not reach a server, reported an error, or did not answer before the job's deadline.

A change never continues because of silence or because time passed. The bake window only delays a decision.

**Before anything changes**, your rollback template runs in check mode on the first wave's servers. If that rehearsal does not pass, the change does not start.