A small endpoint or probe a service exposes so a load balancer or orchestrator knows if it's ready to receive traffic.
A health check is a lightweight endpoint (typically /healthz or /readyz) that returns 200 if the service is healthy and ready for traffic, and a non-2xx code if it isn't. Load balancers, Kubernetes probes, and service meshes use health checks to decide whether to route traffic to a given replica. The two flavors are liveness (am I alive at all, restart me if I'm not) and readiness (am I ready to serve, don't route traffic to me if I'm not).
A service that lies on its health check, returning 200 when it can't actually serve, is the most insidious cause of partial outages: the load balancer keeps sending traffic to it, the dashboards show it as healthy, and customers see errors. Investing in honest health checks (real dependency checks, not just 'process is up') is the single highest-ROI architecture decision for SRE.
See the part of the platform that handles health check in production.