Skip to content
English
Do you need help? Customer portal
neowit.io
  • There are no suggestions because the search field is empty.

Health Monitor - A flap wasn't detected

A subject is visibly bouncing between healthy and unhealthy, but its status never reaches Flapping. Here's how to confirm what the monitor saw and decide what to do. For admins triaging a noisy subject.

How flap detection works

A subject is flagged Flapping when it crosses the flap threshold number of healthy↔unhealthy transitions inside the flap window — currently 5 transitions in 10 minutes (see Policy settings). Both have to be satisfied inside the same rolling window. The values are tuned platform-side by Neowit and aren't per-org.

Step 1 — Read the transitions timeline

Open the subject's detail page on the Health Monitor and count the actual healthy↔unhealthy crossings in the last window:

  • Fewer crossings than the threshold? The subject isn't flapping by the current rule — it's just intermittently unhealthy. Each unhealthy stretch fires its own ticket and auto-resolves on recovery.
  • Enough crossings, but spread over longer than the window? They don't all fall inside one rolling window, so the threshold never trips even though the subject is clearly noisy.

A subject detail page transitions timeline showing four healthy→unhealthy→healthy crossings over twelve minutes — one short of the default five-in-ten-minutes flap rule.

Step 2 — Check whether debounce is absorbing the blips

The open and resolve debounces (currently ~60 seconds each) smooth out brief blips before they become observed state changes. A crossing only counts toward the flap rule once it's observed as a real transition — so if a subject recovers inside the debounce window, that blip never registers and never counts.

This is by design — the debounce exists precisely to suppress single short blips.

Step 3 — Confirm you're looking at the right signal

Flapping is a live status, surfaced on the Health Monitor page. The flap count is also exposed on the workflow trigger as ticket.flapCount, so a workflow can branch on it. If you expected a ticket tagged as flapping and didn't get one, check that the subject isn't silenced — a silence suppresses ticket creation while still updating status.

What you can do

The flap window, threshold, debounces, and post-flap stable window are tuned platform-side and aren't per-org settings. If a subject is genuinely noisy below the flap threshold:

  • Silence the subject during scheduled maintenance windows so it doesn't keep opening tickets (see Silences).
  • Fix the upstream cause. A subject crossing healthy↔unhealthy a few times per hour is reporting something — flaky power, weak gateway coverage, an integration with a noisy heartbeat. The flap rule is meant to catch genuine churn that shouldn't auto-resolve; sub-threshold bouncing is usually a real signal worth investigating at the source.
  • Talk to support if your environment routinely produces flaps that you'd want flagged but the current threshold misses.

Related