Skip to main content

Java Profiling

Overview

light-profiling-java-collect.sh is a shell script used to perform profiling on Java applications. It uses JFR to collect application performance data and upload it to Lighthouse.

Prerequisites

Prerequisites
  1. Install light-agent
  2. Log in to the Lighthouse platform, go to the "Data Collection" module to get light-agent installation instructions and steps.

Environment Variables

Here are some important environment variables, their default values, and purposes:

  • LIGHT_WAY_URL:
    • Default value: http://localhost:8085
    • Purpose: Specifies the machine IP where the Lighthouse platform is located, used for uploading generated JFR files.
  • APP_ENV:
    • Default value: dev
    • Purpose: Specifies the application environment, such as dev, staging, prod, etc.
  • APP_VERSION:
    • Default value: 1.0.0
    • Purpose: Specifies the application version number.
  • HOST_NAME:
    • Default value: Current hostname
    • Purpose: Specifies the hostname to identify the data source.
  • SERVICE_NAME:
    • Default value: Empty
    • Purpose: Specifies the service name to identify the data source.
  • PROFILING_DURATION:
    • Default value: 60 seconds
    • Purpose: Specifies the duration of profiling (in seconds).
  • PROFILING_EVENT:
    • Default value: cpu,alloc,wall,lock
    • Purpose: Specifies the types of profiling events, supported event types include cpu, alloc, wall, lock, live, etc.
  • PROCESS_ID:
    • Default value: Empty
    • Purpose: Specifies the Java process ID(s) to collect from, multiple process IDs can be separated by commas.
  • PROFILING_TAGS:
    • Default value: Empty
    • Purpose: Specifies additional tags for identifying and categorizing data.

How to Execute

  1. Prepare Environment

    Ensure the following tools and libraries are installed:

    • bash
    • curl
    • gzip
    • uuidgen
    • jps (usually installed with JDK)

  2. Set Executable Permission, ensure the script has executable permission.

cd /usr/local/light-agent/profiling/java
chmod +x light-profiling-java-collect.sh
  1. Set Environment Variables Set environment variables as needed. You can set them directly in the script or before running the script.
export LIGHT_WAY_URL=http://your-light-trans-url:8085     
export APP_ENV=prod     
export APP_VERSION=2.0.0     
export HOST_NAME=your-hostname     
export SERVICE_NAME=your-service-name     
export PROFILING_DURATION=120     
export PROFILING_EVENT=cpu,alloc,wall     
export PROCESS_ID=1234,5678     
export PROFILING_TAGS=tag1,tag2
  1. Run Script Run the script to start collecting profiling data.
cd /usr/local/light-agent/profiling/java
nohup ./light-profiling-java-collect.sh &
  1. View Logs The script will generate log files in the runtime directory, one log file per process ID. You can view logs using:
cd /usr/local/light-agent/profiling/java
cat logs/1234.log
  1. Stop Script The script will run in an infinite loop until manually stopped. You can use Ctrl+C to terminate the script.

Notes

  • Log Files: The script will append logs to corresponding files in the runtime directory. If you need to clear log files, you can manually delete or empty these files.
  • Resource Consumption: Running the script for a long time may consume significant system resources. Please adjust PROFILING_DURATION and PROFILING_EVENT according to actual needs.
  • Network Connection: Ensure the LIGHT_WAY_URL service address is reachable, otherwise data upload will fail. The IP is the machine where Lighthouse is installed, default port is 8085.

Common Issues

  • Q: Why are no log files generated?

    A: Ensure the runtime directory exists and has write permissions. If no log files are generated, it might be because the script couldn't find the corresponding Java process.

  • Q: What if data upload fails?

    A: Check if the LIGHT_WAY_URL service address is correct and the network connection is normal. Check the script's error output to locate the issue.

  • Q: How to adjust profiling event types?

    A: Adjust profiling event types by setting the PROFILING_EVENT environment variable. For example: export PROFILING_EVENT=cpu,alloc,wall,lock.