How to Use Arthas monitor to Track Java Method Performance and Latency

What is Arthas?

Arthas is an open‑source Java diagnostic utility that provides real‑time insight into running JVM processes. It offers commands for class inspection, thread analysis, and performance monitoring without requiring application restarts.

Key Command: monitor

The monitor command focuses on tracking the execution of a specific Java method. It periodically records a set of metrics that help you understand how often the method is called, how many calls succeed or fail, and how long each call takes.

Monitoring Metrics

timestamp : The time when the record was taken.

class : Fully qualified Java class name.

method : Method name (including constructors).

total : Total number of invocations.

success : Number of successful executions.

fail : Number of failed executions (exceptions are counted as failures).

rt : Average response time (in milliseconds) of the method.

fail‑rate : Percentage of failed calls.

Among these, the rt (average response time) is especially useful for spotting performance regressions under different input parameters.

Command Options

The command accepts a single option -c that defines the sampling period. If omitted, the default interval is 120s. Longer intervals reduce overhead but provide coarser data.

Performance Impact

Running monitor does introduce some overhead. In a simple experiment, enabling the command increased the average response time of the target method by roughly 50 %. Once Arthas is detached, the JVM returns to its original performance levels.

Example Code

package com.fun;

import com.fun.frame.httpclient.FanLibrary;
import com.fun.utils.Time;

public class AR extends FanLibrary {
    public static void main(String[] args) {
        while (true) {
            sleep(1000);
            long mark = Time.getTimeStamp();
            for (int i = 0; i < 1000; i++) {
                output(DEFAULT_STRING);
            }
            long mark1 = Time.getTimeStamp();
            System.out.println(mark1 - mark);
        }
    }

    static String test() {
        sleep(100);
        return DEFAULT_STRING;
    }

    String fun() {
        sleep(100);
        int randomInt = getRandomInt(2);
        if (randomInt == 1) fail();
        return DEFAULT_STRING;
    }
}

When the above program runs, you can attach Arthas and execute monitor com.fun.AR fun -c 30 to observe the metrics listed earlier.