Multi-region outage verification

When monitoring runs from only one location, a brief routing glitch at our internet provider can wake you at three in the morning with a false alarm. Multi-region verification solves this: an outage is confirmed only when it is reported by the primary server and at least one backup in a different region or at a different internet provider.

How it works

1. The main check (from our central server) reports DOWN (unreachable).
2. Before sending an alert, the system calls a backup worker in another region.
3. The worker runs its own test and usually returns UP or DOWN within a few seconds (depending on the monitor timeout).
4. If the backup also confirms DOWN, the monitor is marked unreachable and alerts are sent.
5. If the primary server reported DOWN but the backup returned UP, it was a false alarm (a flake) - no alert is sent.

When consensus kicks in

Multi-region verification is not triggered on every check - only in clearly defined cases, so it does not slow down routine measurement:

• The primary server first runs its own check, including internal retries (2 attempts with a short pause).
• Consensus is triggered only when the primary still reports DOWN after the retry. If everything is fine, the backup regions are not called at all.
• It applies only to monitor types http (without multi-step flows), tcp and ping. Other types (ssl, dns, domain, visual, heartbeat) do not use it.

How an outage is decided

After the primary reports DOWN, the system requests an independent check from the available backup workers in other regions. From the results it counts how many locations confirmed the outage:

• The monitor is marked unreachable only when the number of DOWN reports reaches the configured threshold (default 2 regions including the primary).
• If the threshold is not reached (for example the primary reported DOWN but the backup reported UP), it is a false alarm - the monitor stays UP and no alert is sent. The response time used is the average of the regions that answered cleanly.
• A backup worker that is not responding at the moment is silently skipped and not counted in the vote.

Who enables it and how

Multi-region verification is an optional platform-wide feature managed by the ePulz.io administrator. The regions are not fixed - the administrator adds backup workers as needed (for example the primary server plus one or two cloud nodes in other countries) and sets how many of them must confirm an outage.

How it appears in the interface

In the monitor detail, for an outage that went through consensus you can see the verification result from each location, for example:

consensus: primary:down, eu1:down
result: DOWN (2 of 2 confirmed)

On the public status page visitors see no technical detail - they only get a verified, trustworthy service status without false alarms.

Practical implications

• Fewer false alerts - a brief glitch on the route to one provider is filtered out by verification from another region.
• Slightly longer detection time - confirming an outage adds a few extra seconds compared to measuring from a single location.
• More reliable history - only real outages, not local network fluctuations, are written into incidents and availability statistics.