Notification throttling

Alert noise is the death of any monitoring system. DevFlow's defaults are tuned to suppress duplicate alerts and respect maintenance windows.

dedup

Once a monitor is in an open incident (incident-management), subsequent failing checks do not fire new pages. The monitor's incident page accumulates the failure count. The on-call still sees the trend in the channel thread, but the pager doesn't ring 60 times.

mute windows

You can mute a monitor by tag:

devflow mute apply --tag service=payments-api --duration 1h --reason "deploy"

Or in YAML for a scheduled deploy:

mute:
  name: payments-deploy-window
  schedule: "0 14 * * THU"  # 2pm UTC every Thursday
  duration: 30m
  selector:
    tags:
      service: payments-api

maintenance windows

Same shape, different intent — maintenance windows are for declared downtime, not noise suppression. They suppress alerts AND reflect on the status-pages public page.

maintenance:
  name: db-failover-q2
  start: 2026-05-12T02:00:00Z
  duration: 2h
  selector:
    tags: { service: payments-api }

quiet hours per channel

Channels can have quiet hours; e.g. SMS-only-during-business-hours, email always. See alert-channels.

anti-pattern: muting because the alert is wrong

If you find yourself muting an alert every Tuesday, the alert is wrong, not the schedule. Fix the assertion (response-assertions) or the retry-policy before reaching for the mute button.