Exposing Metrics of a Spring Boot Application using Dropwizard

How do we know if an application we just put into production is working as it should? How do we know that the application can cope with the number of users and is not slowing down to a crawl? And how do we know how many more users the application can handle before having to scale up and put another instance into the cluster? The answer to these questions is transparency. A good application is transparent in that it exposes several metrics about its health and current status that can be interpreted manually as well as automatically.

This post explains how to create metrics in a Java application with the Dropwizard metrics library and how to expose them with Spring Boot.

What Metrics to Measure?

Usual monitoring setups measure metrics like CPU, RAM and hard drive usage. These metrics measure the resources available to our application. These metrics can usually be read from the application server or operating system so that we don’t have to do anything specific within our application to make them available.

These resource metrics are very important. If the monitoring setup raises an alarm because some resource is almost depleted, we can take action to mitigate that problem (i.e. adding another hard drive or putting another server into the load balancing cluster).

However, there are metrics which are just as important that can only be created within our application: the number of payment transactions or the average duration of a search in a shop application, for example. These metrics give insight to the actual business value of our application and make capacity planning possible when held against the resource metrics.

Creating Metrics with Dropwizard

Luckily, there are tools for creating such metrics, so we don’t have to do it on our own. Dropwizard metrics is such a tool, which makes it very easy to create metrics within our Java application.

Injecting the MetricsRegistry

First off, you will need a MetricRegistry object at which to register the metrics you want to measure. In a Spring Boot application, you simply have to add a dependency to the Dropwizard metrics library. Spring Boot will automatically create a MetricRegistry object for you which you can inject like this:

@Service
public class ImportantBusinessService {
  
  private MetricRegistry metricRegistry;
  
  @Autowired
  public ImportantBusinessService(MetricRegistry metricRegistry){
    this.metricRegistry = metricRegistry;
  }
  
}

Measuring Throughput

If you want to create a throughput metric or a “rate”, simply create a Meter and update it within your business transaction:

@Service
public class ImportantBusinessService {
  
  private Meter paymentsMeter;
  
  @Autowired
  public ImportantBusinessService(MetricRegistry metricRegistry){
    this.paymentsMeter = metricRegistry.meter("payments");
  }
  
  public void pay(){
    ... // do business
    paymentsMeter.mark();  
  }  
  
}

This way, each time a payment transaction is finished, Dropwizard will update the following metrics (also see the Dropwizard manual) :

• a counter telling how many payments have been made since server start
• the mean rate of transactions per second since server start
• moving average rates transactions per second within the last minute, the last 5 minutes and the last 15 minutes.

The moving averages rates are actually exponentially weighted so that the most recent transactions are taken into account more heavily. This is done so that trend changes can be noticed earlier, since they can mean that something is just now happening to our application (a DDOS attack, for example).

Measuring Duration

Dropwizard also allows measuring the duration of our transactions. This is done with a Timer:

@Service
public class ImportantBusinessService {
  
  private Timer paymentsTimer;
  
  @Autowired
  public ImportantBusinessService(MetricRegistry metricRegistry){
    this.paymentsTimer = metricRegistry.timer("payments");
  }
  
  public void pay(){
    Timer.Context timer = paymentsTimer.time();
    try {
      ... // do business
    } finally {
      timer.stop();
    }
  }  
  
}

A Timer creates the following metrics for us:

• the min, max, mean and median duration of transactions
• the standard deviation of the duration of transactions
• the 75th, 95th, 98th, 99th and 999th percentile of the transaction duration

The 99th percentile means that 99% of the measured transactions were faster than this value and 1% was slower. Additionally, a Timer also creates all metrics of a Meter.

Exposing Metrics via Spring Boot Actuator

Having measured the metrics, we still need to expose them, so that some monitoring tool can pick them up. Using Spring Boot, you can simply add a dependency to the Actuator Plugin. By default Actuator will create a REST endpoint on /metrics which lists several metrics already, including some resources metrics as well as counts on different page hits.

Spring Boot has support for Dropwizard by default, so that all metrics created with Dropwizard will automatically be exposed on that endpoint. Calling the endpoint results in a JSON structure like the following:

{  
  "classes": 13387,
  "classes.loaded": 13387,
  "classes.unloaded": 0,
  "datasource.primary.active": 0,
  "datasource.primary.usage": 0.0,
  "gc.ps_marksweep.count": 4,
  "gc.ps_marksweep.time": 498,
  "gc.ps_scavenge.count": 17,
  "gc.ps_scavenge.time": 305,
  "heap": 1860608,
  "heap.committed": 876544,
  "heap.init": 131072,
  "heap.used": 232289,
  "httpsessions.active": 0,
  "httpsessions.max": -1,
  "instance.uptime": 3104,
  "mem": 988191,
  "mem.free": 644254,
  "nonheap": 0,
  "nonheap.committed": 115008,
  "nonheap.init": 2496,
  "nonheap.used": 111648,
  "processors": 8,
  "systemload.average": -1.0,
  "threads": 19,
  "threads.daemon": 16,
  "threads.peak": 20,
  "threads.totalStarted": 25,
  "uptime": 20126,
  "payments.count": 0,
  "payments.fifteenMinuteRate": 0.0,
  "payments.fiveMinuteRate": 0.0,
  "payments.meanRate": 0.0,
  "payments.oneMinuteRate": 0.0,
  "payments.snapshot.75thPercentile": 0,
  "payments.snapshot.95thPercentile": 0,
  "payments.snapshot.98thPercentile": 0,
  "payments.snapshot.999thPercentile": 0,
  "payments.snapshot.99thPercentile": 0,
  "payments.snapshot.max": 0,
  "payments.snapshot.mean": 0,
  "payments.snapshot.median": 0,
  "payments.snapshot.min": 0,
  "payments.snapshot.stdDev": 0
}

Wrap-Up

When implementing a web application, think of the business metrics you want to measure and add a Dropwizard Meter or Timer to create those metrics. It’s a few lines of code that provide a huge amount of insight into an application running in production. Spring Boot offers first class support for Dropwizard metrics by automatically exposing them via the ‘/metrics’ endpoint to be picked up by a monitoring tool.