Micrometer and Prometheus to monitor Spring Boot application

Prometheus is a time-series database - and suitable for storing application metrics.

Micrometer - as defined by their dev team - is like SLF4J, but for metrics.

We can use a combination of micrometer and Prometheus to monitor spring boot applications. Well, we need a visualization tool like Grafana to make more sense of metrics, but let us leave it for the next article.

Setup

As mentioned, we shall set up 3 things

1. Spring Boot application
2. Micrometer to expose actuator metrics in Prometheus format
3. Prometheus to store and make sense of metrics

Setup SpringBoot application

start.spring.io is the best place to get started. 

But do remember to add actuator dependency.

Here is what I did as an example:

Create an endpoint that you would like to play with. I added a chance endpoint that looks like this:

@RestController

@RequestMapping("/chance")

public class ChanceController {

  @GetMapping

  public ResponseEntity chances() {

   double dChance = Math.random() * 10;

   int chance = (int) dChance;

   System.out.println("Chance: " + dChance);

   if(chance <= 7) {

     return new ResponseEntity("Success", HttpStatus.OK);

   } else if(chance == 8) {

     return new ResponseEntity("Failure", HttpStatus.UNAUTHORIZED);

   } else {

     return new ResponseEntity("Failure", HttpStatus.INTERNAL_SERVER_ERROR);

   }

  }

}

This endpoint has an 80% chance of success, 10% chance of 401 (Unauthorized) response, and 10% chance of 500 (Internal Server Error) response.

Now our spring boot application is good to go.

Check the endpoints exposed at localhost:8080/actuator.

Setup micrometer

Micrometer can be thought of as middleware or an adapter. It converts Spring Boot's actuator metrics in a format that can be read by any metric collection tool. Prometheus is one such tool it supports. For a complete list, refer to this documentation.

So we add a dependency to micrometer-registry-prometheus. e.g.

<dependency>

    <groupId>io.micrometer</groupId>

    <artifactId>micrometer-registry-prometheus</artifactId>

</dependency>

Now restart the spring boot app. A new endpoint is created that exposes metrics in a format that Prometheus can understand. This endpoint is http://localhost:8080/actuator/prometheus.

Setup Prometheus

We setup Prometheus as a docker container in our previous article Prometheus on Docker. Just use the commands in the Run section to set up Prometheus on docker.

Update prometheus.yml file to look like this:

global:

  external_labels:

    monitor: 'codelab-monitor'

scrape_configs:

  - job_name: 'spring_boot_project'

    scrape_interval: 5s

    metrics_path: '/actuator/prometheus'

    static_configs:

      - targets: ['docker.for.mac.host.internal:8080']

Notice the targets. Ideally, I would have used 'localhost'. However 'localhost' would refer to the docker container. I wish to point to my laptop. On mac, you can do this as 'docker.for.mac.host.internal'. On windows, you can do this as 'host.docker.internal'. In summary - refer to your docker documentation. Or stackoverflow.

Also 'metrics_path' points to the new actuator endpoint added by the multimeter.

View metrics

Open Prometheus on localhost:9090 and go to Graph.

Use this as a query. Replace '/chance' with the endpoint you wish to visualize.

http_server_requests_seconds_count{method="GET", uri="/chance"}

My graph looks like this:

Explaining prometheus query language and usage of Prometheus is beyond scope of this article.

Conclusion

Spring Boot metrics can be collected by multimeter and Prometheus sufficiently well.

In my experience, APMs like DataDog and SignalFx are very well suited for application monitoring. I loved DataDog and its simplicity. Although SignalFx is good enough too. However, these are costly APMs and can drain much-needed money from a startup or a small company. So, for a start, multimeter + Prometheus + Grafana are a good start.