Quorum and regions
The regions
Section titled “The regions”Six probe regions, one per inhabited continent: na-ewr, sa-gru, eu-fra, af-jnb, as-sgp, oce-syd. Free probes from 2 regions, Pulse from 3, Sentinel and up from all six. The network page covers the fleet architecture.
The voting model applies to the eleven regional probe types. Agent and heartbeat monitors report their signals directly and follow their own incident rules.
Two thresholds before an alert
Section titled “Two thresholds before an alert”- Within a region, the algorithm folds fresh node results and requires 75% agreement when more than one probe node is active. Today each of the six regions runs one active node, so this layer currently resolves as 1/1, and the confirmation comes from across regions. Additional nodes per region are planned, and this layer then adds a second confirmation inside each region.
- Across regions, an incident opens when a quorum of regions confirms the failure, two regions by default. Recovery has its own quorum, and only results inside the freshness window count. This cross-region vote carries the confirmation in today’s topology.
The incident’s start time is the quorum confirmation, to the second, and the recovery boundary is the confirmed recovery. Those are the exact window boundaries your availability record uses. Nothing is rounded, nothing is backdated.
Below quorum: forming and watchlist
Section titled “Below quorum: forming and watchlist”- A forming incident (first region failing, quorum not yet reached) is visible in the product with the first failing region named, and sends exactly one quiet notification per episode.
- Sub-quorum divergences (one region drifting or one IP family failing) land on a watchlist. Node disagreement also appears there when a region has more than one active node. An entry must persist for a full check interval before it is reported, so you investigate on your schedule, not at three in the morning.
One explicit rule for alerts and evidence
Section titled “One explicit rule for alerts and evidence”The default cross-region quorum is two, and an alert profile can change that policy. The record keeps the configured rule beside each incident boundary. For the eleven regional types, the same explicit policy that filters out single-path alerts also defines the availability window, so the alarm and the evidence always agree. The rule confirms that a service failed, and the incident timeline and your team establish why.
- From measurement to alarm: the full path from a check to a ringing phone.
- Evidence and retention: how these boundaries become SLA figures.