Capture Java application metrics

Inspired by Metrics written by Coda Hale. But don't want to get too complicated. I want something lightweight but just enough.

Minimum overhead on data collecting. Use thread to emit event when time unit reached.

Depend on listener implementation, data can be send to database, in-memory, another service, or wherever useful.

0.1.2

Start point. Singleton class. Default time unit is 1 minute. Tracker.count(key); Tracker.incr(key); Tracker.decr(); Tracker.gauge(key, value); Tracker.log(key, log);

Tracker.peek(); // Emit onTimeUnit with current data.

Base class

If aspect is set to accumulative then values do not reset on each time unit. Default is non-accumulative.

Count: Counts number per time unit.

Incr/Decr: Keep counts over time. Does not reset number if the counter is used with incr/decr.

Collect dataset per time unit. Calculates count/total/main/mean/90 percentile/95 percentile/99 percentile/max when emitting event.

Keep special logs per one time unit.

Implements IListener interface. The method onTimeUnit will be triggered on every time unit with collected data in JSON string.

Application should implement IListener and add to Tracker to deal with data.

public interface IListener {
    public void onTimeUnit(String data);
}

Listen to data and writes data to console.

Listen to data and POST to another service.

import com.rz.metrics.Tracker;
import com.rz.metrics.listeners.ConsoleListener;

Tracker.setTimeUnit(1000L); // 1 second
Tracker.addListener(new ConsoleListener());

Tracker.count("traffic:url1");

If first time calling a counter, which creates the counter, is by incr/decr then the counter is set to accumulative automatically.

Tracker.incr("onlineUser"); // Login
Tracker.decr("onlineUser"); // Logout or expire

long latency = nnn;
Tracker.gauge("latency:url1", latency);

Tracker.log("malicious", "IP xxx.xxx.xxx.xxx");

/**
 * @param data
 *         {
 *             aspect: "counter"|"gauger"|"logger"
 *             , key: "..."
 *             , ts: "nnn" // Timestamp
 *             , timeUnit: "60000"
 *             , data: {...}
 *         }
 */
public void onTimeUnit(String data) {
    // Do whatever.
}

java -jar metrics-0.1.0.jar

cd nodeClient
node app.js

Open browser, goto http://localhost:3001/index.html