Micrometer Metrics: Counters, Timers, and Gauges in Spring Boot

Micrometer provides three primary instrument types for Spring Boot metrics — Counter, Timer, and Gauge — and choosing the right one matters because the wrong type produces dashboards that can't answer the questions you need to ask.

Counter — for things that only go up

• A Counter tracks a cumulative count that only ever increases — total orders placed, total payment failures, total cache misses; you never reset a Counter
• In dashboards, you query the rate of change (rate(counter[5m]) in Prometheus) not the raw value — which means a Counter of 10,000 orders tells you nothing until you ask "how many orders per minute in the last 5 minutes?"
• Counter.builder("orders.placed").tag("status", status).register(meterRegistry).increment()

Timer — for duration and count together

• A Timer records both the duration of each operation and the count of operations — which means one Timer instrument gives you request rate, average latency, and latency percentiles (P50, P95, P99) all in one
• Timer.builder("payment.processing").register(meterRegistry).record(() -> paymentService.process(payment)) wraps the call and records its duration automatically
• Spring Boot's Actuator auto-creates a Timer for every HTTP endpoint — http.server.requests tagged with uri, method, and status; you get latency distributions without writing any code

Gauge — for values that go up and down

• A Gauge reports the current value of something that fluctuates — queue depth, active connections, cache size, JVM heap percentage
• Gauge.builder("queue.depth", queue, Queue::size).register(meterRegistry) reports the current queue size every time Micrometer scrapes the metric
• Use Gauge for current state; use Counter for cumulative counts; never use Gauge for things that only go up (use Counter for those)

Tags create dimensions — tag("service", "payments") and tag("env", "prod") let dashboards filter and group metrics without separate namespaces.