JFR playground

A proof of concept of how to collect metrics using the JDK Flight Recorder event streaming(JEP 349) and Kamon telemetry to send them to Prometheus/Pinot

Prerequisites

Java 17+
sbt
k6

Kamon + Prometheus + Grafana

# Prometheus + Grafana + Metrics dashboard
docker-compose -f docker/grafana-prometheus-compose.yml up

# Example service
sbt run

# Stress the service
k6 run k6/stress.js      

          /\      |‾‾| /‾‾/   /‾‾/   
     /\  /  \     |  |/  /   /  /    
    /  \/    \    |     (   /   ‾‾\  
   /          \   |  |\  \ |  (‾)  | 
  / __________ \  |__| \__\ \_____/ .io

  execution: local
     script: k6/stress.js
     output: ...........

Then open the "JFR Metrics" dashboard in Grafana(localhost:3000) and Enjoy!

Bonus

Send profiling information(jdk.ObjectAllocationSample) to Pinot realtime table via Kafka and play with it ;)

# Kafka + Pinot
docker-compose -f pinot-kafka-compose.yml up

# Topic
docker exec -t kamon-jrf-playground_kafka  kafka-topics --create --bootstrap-server localhost:9092 --replication-factor 1 --partitions 1 --topic jfr-allocation-topic

# Table Schema
curl -X POST "http://localhost:9000/schemas?override=true" -H "accept: application/json" -H "Content-Type: application/json" -d @jfr-allocation-schema.json

# Realtime Table configuration
curl -X POST "http://localhost:9000/tables" -H "accept: application/json" -H "Content-Type: application/json" -d @jfr-object-allocation-table.json

# Enable `jdk.ObjectAllocationSample` event in application.conf changing
"jdk.ObjectAllocationSample#enabled=false" to "jdk.ObjectAllocationSample#enabled=true"

Then open the Pinot SQL Console(localhost:9000/#/query) and Enjoy!

Disclaimer

I have just scratched the surface of Pinot capabilities, and I'm sure there are other better ways of modeling than the one proposed, using other features and data types.

License