Learn › Exchange Online › 10 · Operations & troubleshooting

Mail-flow monitoring and alerting

Exchange shows no dials by default — you build them. The EXO signal set for the four-class model: invisible queues, deliverability you construct, and the auth-failure canaries that catch attacks and retirements early.

The EXO signal set (four classes, mail edition)

  1. Service health: Exchange advisories (scan leg) — 'is it them', first.
  2. Quality/deliverability: the BUILT telemetry — latency percentiles (trace concept), NDR-reason trends (ledger), outbound-reputation smoke (5.7.7xx + restricted-sender events), deferral trends. Exchange gives you nothing here out of the box; you construct it from traces + reports.
  3. Config drift: the scanned 40+ types diffed — the posture-drift watchlist, alerting.
  4. Change-from-Microsoft: MC posts verdicted against your EXO config (EWS/SMTP-AUTH/POP clocks) — the site's core loop.

The canaries worth building

  • Auth-failure canary: legacy-auth / basic-SMTP failure trends — catches BOTH the attack (credential stuffing) and the retirement casualty (December's doomed scanner) in one signal.
  • Delivery canary: scheduled test-mails through each path with trace verification — the SBA-drill logic for mail; advisory-beating detection of quiet path breaks.
  • The persistence trio (BEC): inbox-rule-create / forwarding-change / oauth-grant — security alerts that are also mail-flow events.

Routing

A PAGE for 'how's mail flowing', a PING only for 'act now' (reputation smoke, the trio, a canary failing, a watchlist drift) — to the owner who can act. Mail generates a lot of noise; disciplining it is the whole game.

What to watch (proofs)

  • Coverage honesty: which classes you actually monitor — most estates have (1) by accident; the gap is the backlog.
  • Canary liveness: last-green timestamps — a dead canary monitors nothing.
  • Alert-to-action ratio per signal — the pruning and fatigue metric.
  • The correlation payoff: a deliverability dip against drift + MC verdicts on one timeline (Pulse) — mail's root-cause-in-minutes.

PowerShell for this concept

Discussion

No messages yet — start the thread.

Sign in with your email to join the discussion — we send a one-time link, no password.