How to Instantly Diagnose Java Production Issues with Arthas

Introduction

Before using Arthas, diagnosing Java production issues required many commands (jps, jstack, jmap, etc.) which was cumbersome. Arthas simplifies locating problems such as CPU spikes, high load, memory leaks, allowing quick resolution.

1. What is Arthas

Arthas is an open‑source Java diagnostic tool released by Alibaba in September 2018. It supports JDK 6+ and provides an interactive command‑line interface with tab completion. The project has over 17 000 stars on GitHub.

Official documentation and issue examples are available on the GitHub repository.

Open‑source address: https://github.com/alibaba/arthas

Official docs: https://alibaba.github.io/arthas

2. Typical Use Cases

Get a global view of system status.

Identify which class or method is consuming CPU.

Detect deadlocks or blocked threads.

Measure method execution time.

Find the JAR that loaded a class.

Verify whether recent code changes are taking effect.

Debug without adding logs or redeploying.

Monitor JVM runtime metrics in real time.

3. How to Use Arthas

3.1 Installation

Download the arthas‑boot.jar from the official GitHub or Gitee mirror.

# github download
wget https://alibaba.github.io/arthas/arthas-boot.jar
# or Gitee download
wget https://arthas.gitee.io/arthas-boot.jar
# print help
java -jar arthas-boot.jar -h

3.2 Running

Start the jar and select a Java process, or pass the PID directly.

# start and then choose PID
java -jar arthas-boot.jar
# or start with PID
java -jar arthas-boot.jar <pid>

Use ps or jps to list Java processes.

After the Arthas logo appears you can issue diagnostic commands.

3.3 Web Console

Arthas also provides a Web Console accessible at http://127.0.0.1:8563/ with the same functionality as the terminal.

3.4 Common Commands

Some frequently used commands include:

dashboard – real‑time system overview.

thread – list thread information and CPU usage.

watch – observe method parameters and return values.

trace – show call stack and time spent per method.

stack – display the call chain of a method.

tt – time‑travel tunnel for method invocations.

monitor – periodic method execution statistics.

sc – view loaded class information.

sm – list methods of a class.

jad – decompile a loaded class.

classloader – inspect classloader hierarchy.

heapdump – generate a heap dump similar to jmap.

4. Practical Scenarios

4.1 Global Monitoring

Use dashboard to see threads, memory, GC, and environment data.

4.2 CPU Spike Diagnosis

Run thread to list all threads and their CPU usage; identify the thread with high CPU (e.g., ID 12).

Drill down with thread 12 or thread -n 10 to see top‑N CPU threads.

4.3 Thread‑Pool Status

Inspect thread‑pool threads with thread | grep pool to see WAITING threads.

4.4 Deadlock Detection

Run thread -b to display deadlock information.

4.5 Decompilation

Use jad com.Arthas to view source code of a loaded class.

4.6 Field and Method Inspection

Command sc -d -f com.Arthas shows fields; sm com.Arthas lists methods.

4.7 Variable Observation

Use ognl '@com.Arthas@hashSet' to inspect static variables, size, and modify them.

4.8 Performance Tracing

Apply trace com.UserServiceImpl get to see time spent in each sub‑method, identifying the most expensive part (e.g., mysql()).

4.9 Method Monitoring

Run monitor -c 5 com.UserServiceImpl get to collect execution statistics every 5 seconds.

4.10 Parameter Watching

Command watch com.Arthas addHashSet '{params[0],returnObj}' prints input and output of a method.

4.11 Call Stack

Use stack com.UserServiceImpl mysql to view the call chain.

4.12 Time‑Travel Tunnel

Start recording with tt -t com.UserServiceImpl check, list records with tt -l, view details with tt -i 1001, and replay a specific invocation using tt -i 1001 -p.

These commands enable developers to locate, analyze, and fix production issues without modifying code or restarting services.