Multi-region monitoring

Single-region monitors lie. They can't tell the difference between a global outage and a Virginia-specific routing issue.

Multi-region monitors run from N regions in parallel. The status only flips to down if every region fails. If only some fail, the monitor is degraded and the alert lists which regions saw the failure.

Picking regions

Default for new monitors: ["us-east"]. Update via regions: [...] on create or update.

Region codeDatacenter
us-eastAshburn, VA
us-westSan Jose, CA
eu-westLondon, UK
eu-centralFrankfurt, DE
ap-southeastSingapore
ap-northeastTokyo, JP

Free tier: max 3 regions per monitor. Pro+: all 6.

Aggregation

For each round of checks, the per-region results aggregate:

Per-regionAggregated
All upup
1+ down or degraded, not alldegraded
All downdown

Tuning to your audience

Use all 6. The cost is overlap (you'll see brief regional glitches as degraded), but you catch every regional outage.

us-east + us-west is plenty. Adding EU/AP regions creates noise from network glitches in regions you don't serve.

Multi-region won't work — only one region has VPN access. Use heartbeat monitoring for self-reported health, or skip multi-region entirely.

Use eu-west + eu-central only. No US/AP probes touch your endpoint.

Reducing flap

Multi-region monitors will sometimes flicker between up and degraded due to short-lived network blips in one region. Two ways to reduce noise:

  1. Increase confirmation_period_s on the monitor — wait N seconds of consistent results before flipping state.
  2. Drop regions you don't really need.
Ask a question... ⌘I