Check if there are JVM metrics we can automatically track

[adinauer]

Check if there are JVM metrics we can automatically track and send them to Sentry using Metrics API.

Opt-in

things that could be interesting: CPU, memory, GC

some metrics might require specific JVM flags, JFR, etc.

[linear]

JAVA-253

[adinauer]

Auto-Collected Metrics Ideas (Java + Spring Boot)

This document lists candidate metrics the Java SDK could collect automatically using
IMetricsApi for common Java and Spring Boot workloads. Each metric includes a
suggested type and typical low-cardinality attributes.

A. JVM/Runtime

Counts and gauges based on JVM, OS, and process signals.

• jvm.gc.count (count): total GC cycles by collector.
• jvm.gc.time (distribution, ms): GC pause duration by collector.
• jvm.heap.used (gauge, bytes): heap used (pre/post GC).
• jvm.heap.committed (gauge, bytes): committed heap.
• jvm.heap.max (gauge, bytes): max heap size.
• jvm.nonheap.used (gauge, bytes): metaspace and code cache usage.
• jvm.threads.live (gauge, count): current live threads.
• jvm.threads.daemon (gauge, count): current daemon threads.
• jvm.threads.peak (gauge, count): peak thread count since start.
• jvm.classes.loaded (gauge, count): loaded class count.
• process.cpu.utilization (gauge, ratio): process CPU usage.
• process.uptime (gauge, ms): process uptime.

Typical attributes: runtime (java), collector (gc name), memory_pool (if used).

B. Spring MVC / Spring Boot HTTP Server

Request/response metrics for inbound HTTP server traffic.

• http.server.requests (count): total requests.
• http.server.duration (distribution, ms): request latency.
• http.server.active_requests (gauge, count): inflight requests.
• http.server.errors (count): 5xx responses or unhandled exceptions.
• http.server.response.size (distribution, bytes): response body size.

Typical attributes: http.method, http.route (templated), http.status_code,
server.address, server.port, network.protocol.name, network.protocol.version.

C. Database Clients (JDBC/Hikari, etc.)

Metrics for SQL execution and connection pooling.

• db.client.requests (count): total query executions.
• db.client.duration (distribution, ms): query latency.
• db.client.errors (count): query errors.
• db.client.rows (distribution, count): rows returned/affected (when known).
• db.pool.active (gauge, count): active connections (pool-specific).
• db.pool.idle (gauge, count): idle connections (pool-specific).
• db.pool.pending (gauge, count): threads awaiting connection.
• db.pool.timeout (count): connection acquisition timeouts.

Typical attributes: db.system (postgresql, mysql, etc), db.name, db.operation
(select/insert/update/delete), db.pool.name.

D. Messaging (Kafka, RabbitMQ, JMS)

Publisher/consumer throughput and processing metrics.

• messaging.publish (count): messages produced.
• messaging.publish.duration (distribution, ms): time to publish/ack.
• messaging.consume (count): messages consumed.
• messaging.consume.duration (distribution, ms): processing time per message.
• messaging.errors (count): publish/consume errors.
• messaging.batch.size (distribution, count): batch size (when batching).
• messaging.backlog (gauge, count): lag or queue depth (when observable).

Typical attributes: messaging.system (kafka, rabbitmq, jms),
messaging.destination (topic/queue name), messaging.operation (send/receive),
messaging.destination_kind (topic/queue).