JProfiler Get rid of your performance problems and memory leaks! Modern Garbage Collection Tuning Java Performance Training Courses COURSES AVAILABLE NOW. We can provide training courses to handle all your Java performance needs Java Performance Tuning, 2nd ed The classic and most comprehensive book on tuning Java Java Performance Tuning Newsletter Your source of Java performance news. Subscribe now! Enter email: Training online Threading Essentials course JProfiler Get rid of your performance problems and memory leaks!

Back to newsletter 246 contents

Health checks, metrics, telemetry, logs, tracing, profiling, ... we have a gamut of different things in Observability. How do they all fit together? Here's my take, in order of increasing sophistication:

1. Is the process running? Use a liveness check (a health check). The answer is a yes or no. If no, need to (re)start the process
2. Is the process ready to process requests? Use a readiness check (a more sophisticated health check). The answer is a yes or no. If yes, can route requests to the process, otherwise wait until it is ready, or a timeout is reached (and try restarting it or send an error notification)
3. Is the service performing adequately (adequate performance is defined by a Service Level Objective - SLO)? The answer is a yes or no for each SLO, measured using telemetry (gathering metrics from components) and thresholds. The metrics used to measure the SLO are termed SLIs - Service Level Indicators. Can be measured from every request or from sampled requests; Can be measured at the server-side (service times), or from the actual clients (real-user-metrics), or from client-side generated requests (synthetic transactions).
4. Where is the service failing the SLO? The answer is a prioritized list of hardware-resource/infrastructure-component/JVM-space/application/request/context/process/thread/code stack/line of code (the more sophisticated the tool, the more detailed the answer). Identified by distributed tracing (which needs a unique request ID propagated across services, with transaction context)
5. Why is it failing the SLO? The answer is "because it was doing X", identified by thread-traces/(execution/memory/contention)-profiles/log-events (metrics extracted from logs)/log-lines (most easily found with trace IDs from the previous answer, or by making the logs queryable).

That's about as succinct as you can get in detailing Observability. Now on to all the usual newsletter list of links, tips, tools, news and articles, and as usual I've extracted all the tips into this month's tips page

A note from this newsletter's sponsor

News

Java performance tuning related news

• Java threads use a minimum memory on current JVMs, use -Xss to minimize that if needed
• Improving the Ideal Graph Visualizer for better comprehension of Java's main JIT compiler
• JDK Flight Recorder, The Programmatic Way

Tools

Java performance tuning related tools

• The Ideal Graph Visualizer is a tool developed to help examine the intermediate representation of the JVM C2 compiler
• How to trace jvm filesystem access using Java
• Eclipse MAT: Tidbits - Heap Size, Keep Unreachable Objects, Smart Data Display

Articles

• Demystifying Garbage Collection in Java (Igor Braga/CoffeeJUG)
• Profiling can be about time.. but which time? (Sadiq Jaffer, Richard Warburton/Opsian)
• Object Recycling Techniques on GC and Application Performance (Rafal Foltynski/JVM Roundabout)

Jack Shirazi

Back to newsletter 246 contents