Skip to content

Your first week

  1. Week 1 — observe. Read the incidents. You are checking whether the diagnoses are right, and whether the thresholds match how your host actually behaves. Tune DOCTOR_CPU_THRESHOLD, DOCTOR_LOG_ERROR_BURST and DOCTOR_POLL_INTERVAL here.
  2. Week 2 — suggest. It now proposes actions. Approve them by hand from the dashboard and see whether the proposals were the right call.
  3. Then — auto, selectively. Rather than flipping the whole host, leave the global mode at suggest and opt individual containers in with doctor.mode=auto on the ones where an unattended restart is genuinely safe.