Custom Actuator Endpoint and Prometheus Exposition

Q: Custom Actuator Endpoint and Prometheus Exposition

Create a custom Actuator endpoint with `@Endpoint(id = "name")` and methods annotated `@ReadOperation`, `@WriteOperation`, or `@DeleteOperation` — Spring Boot exposes them at `/actuator/name` automatically. Register the bean and add the ID to `management.endpoints.web.exposure.include`. For Prometheus, add `micrometer-registry-prometheus` and Spring Boot auto-configures `/actuator/prometheus` returning OpenMetrics-format text. Inject `MeterRegistry` to publish custom metrics: counters, gauges, timers, distribution summaries. Always tag metrics with **bounded label sets** — high-cardinality tags (user IDs, request IDs) explode Prometheus storage and become operationally unmanageable.

Intermediate·2 min read·

NetflixDatadogGrafana LabsCloudflareDoorDash

Fast recall

Quick revision

Core answer

Create a custom Actuator endpoint with @Endpoint(id = "name") and methods annotated @ReadOperation, @WriteOperation, or @DeleteOperation — Spring Boot exposes them at /actuator/name automatically. Register the bean and add the ID to management.endpoints.web.exposure.include. For Prometheus, add micrometer-registry-prometheus and Spring Boot auto-configures /actuator/prometheus returning OpenMetrics-format text. Inject MeterRegistry to publish custom metrics: counters, gauges, timers, distribution summaries. Always tag metrics with bounded label sets — high-cardinality tags (user IDs, request IDs) explode Prometheus storage and become operationally unmanageable.

Interview-ready guide

Interview answer

Spring Boot Actuator has two extension points: custom endpoints for app-specific operational data and custom metrics through Micrometer for everything that's quantitative.

Custom endpoints

Annotate a Spring bean with @Endpoint(id = "features")
Add @ReadOperation, @WriteOperation, @DeleteOperation methods
@Selector captures path variables: /actuator/features/{name}
Register by adding the ID to management.endpoints.web.exposure.include
Same Spring Security rules apply as built-in endpoints

Custom metrics

Inject MeterRegistry; create Counter, Gauge, Timer, or DistributionSummary
Use @Timed for declarative method timing — needs a TimedAspect bean
Add micrometer-registry-prometheus and Spring Boot auto-configures /actuator/prometheus
Output is OpenMetrics text; Prometheus scrapes it on a configured interval

Cardinality is the trap

Never tag metrics with unbounded values: user ID, request ID, URL path with IDs
A Counter tagged with user ID becomes one time series per user — millions of series, Prometheus OOM
Use MeterFilter.maximumAllowableTags to cap tag cardinality
Bounded tags are fine: region, HTTP method, status code, error class

In an interview, I'd close with the operational rule: measure first, instrument second. Spring Boot already publishes HTTP latency, JVM memory, JDBC pool stats, and Tomcat thread counts through Micrometer — no custom code needed. Add custom metrics only for KPIs the framework doesn't already cover.

03 · Deep dive