The Kubernetes mechanism that asks each pod two questions: are you alive, and are you ready for traffic.
In Kubernetes, a probe is a periodic check the kubelet runs against a pod. The two main flavors are liveness (if this fails, restart the container, useful for catching deadlocked processes) and readiness (if this fails, stop sending traffic to this pod but don't restart it, useful for warming caches or waiting on dependencies). A third, startup probe, gives slow-booting apps room to come up before the others kick in. Each probe is configured with an HTTP path, a TCP port, or an exec command, plus thresholds.
Misconfigured probes are one of the top causes of Kubernetes outages: a too-strict liveness restarts pods in a loop, a too-loose readiness routes traffic to a pod that can't actually serve. Honest probes that exercise real dependencies (the DB connection, not just 'process is up') are the operational equivalent of testing the smoke alarm, easy to neglect, costly when you need it.
See the part of the platform that handles probe (liveness / readiness) in production.