Spring Boot Actuator turns your application into something you can interrogate at runtime — health, metrics, configuration, threads, loggers — over HTTP or JMX, without writing the plumbing yourself. It's the production-readiness half of Boot: auto-configuration gets the app running; Actuator tells you whether it's healthy once it is. Add spring-boot-starter-actuator and a set of endpoints appears under /actuator.
management:
endpoints:
web:
exposure:
include: health,info,metrics,prometheus
endpoint:
health:
show-details: when-authorizedThe critical default to know: only /health and /info are exposed over HTTP out of the box. Everything else — /metrics, /env, /loggers, /threaddump, /heapdump, /prometheus — must be opted in via management.endpoints.web.exposure.include. That default is deliberate: /env and /heapdump leak secrets and memory, so you never expose them blindly.
Observability rests on three pillars, and Actuator wires up all three. Metrics run through Micrometer, a vendor-neutral facade (think SLF4J, but for meters) that publishes to Prometheus, Datadog, CloudWatch, and others. Health is assembled from HealthIndicator beans — Boot ships ones for the datasource, disk, Redis, etc., and you add your own. Tracing uses Micrometer Tracing (which replaced Spring Cloud Sleuth in Boot 3), bridging to OpenTelemetry or Brave/Zipkin to propagate trace IDs across services.
In Kubernetes, the health endpoint splits into liveness (/actuator/health/liveness) and readiness (/actuator/health/readiness) probes, driven by Spring's AvailabilityState. A failing liveness probe restarts the pod; a failing readiness probe pulls it from the load balancer — two very different remediations.
The questions below cover the endpoint catalogue, custom health indicators, Micrometer's meter types, distributed tracing, locking endpoints down for production, and the liveness-versus-readiness distinction that trips people up in Kubernetes.