The autopilot: acting within policy
The autopilot uses the same checks as a named approval. It can do nothing a person could not approve.
AvailabilityAvailable in early access when your policy allows it. Today the autopilot covers two fixes.
Steps
- A written policy names each cause the autopilot may fix and the one action it may use.
- It acts only on incidents the watch service opened, up to 30 minutes old, at or above the confidence the rule sets, and at most twice an hour.
- Before acting it reads the live state again. After acting it checks the fix worked.
- Every decision is recorded: resolved, did not recover, refused, or handed to a person, with the policy that allowed it.
- Your Teams channel is told every outcome.
What you should see: Today the policy in early access covers two faults: a paused connector and a task still failed after an outage. Each passed its own closed-loop trials before it was switched on.
What it can and cannot do
Can
- Fix the two faults your policy covers without waiting, at most twice an hour, and tell Teams every outcome.
Cannot
- Act on a cause with no rule, below the rule's confidence, past the hourly limit, or while the kill switch or a silence is on.
If it does not work
- A cause with no rule, too little confidence, or the hourly limit reached goes to a person. Nothing is done.
Related articles
- The Policies viewWhat Varqeni is allowed to do: the autonomy level, the autopilot policy, the playbook and every action it knows.
- Stop every action at onceThe kill switch stops all fixes, by a person or the autopilot, immediately.
- Change the autopilot policyThe autopilot policy is a reviewed file on the host. Changes are checked before they take effect.
Still stuck? Contact support or email info@fortisdatagroup.com.