Dead-Man Heartbeat

A "dead man's switch" only works because the check is on the absence side, watched by something other than the thing it's checking. The heartbeat emitter is the push half of that — your application pings an external monitor on a schedule, and the monitor, not your application, decides what "the ping stopped arriving" means.

interface HeartbeatEmitterInterface
{
    /** Send one ping. Returns true if the external monitor acknowledged it. */
    public function emit(HeartbeatPing $ping): bool;
}

emit() must be bounded by a hard timeout and must never block or throw into whatever scheduler loop is calling it — a heartbeat mechanism that can itself hang defeats its own purpose.

php bin/console vortos:observability:heartbeat

Run this on a schedule (cron, a sidecar loop) pointed at any off-host dead-man-switch service. HttpHeartbeatEmitter is the shipped driver — a simple HTTP ping to a configured URL, since most dead-man-switch services (Healthchecks.io-style, BetterStack's heartbeat monitors, your own) expect exactly that.

How this fits the bigger picture

Health's DetectorIndependenceDoctorCheck treats this heartbeat as one of three independent failure detectors a production deploy is required to have configured — alongside in-app probes and an external synthetic uptime monitor. Health's health:monitor:tick command drives this heartbeat directly from the local probe rollup, deciding whether to send a Start, Success, or Fail ping based on what the probes found.