55. Metrics

Spring Boot Actuator provides dependency management and auto-configuration for Micrometer, an application metrics facade that supports numerous monitoring systems, including:

[Tip]Tip

To learn more about Micrometer’s capabilities, please refer to its reference documentation, in particular the concepts section.

55.1 Getting started

Spring Boot auto-configures a composite MeterRegistry and adds a registry to the composite for each of the supported implementations that it finds on the classpath. Having a dependency on micrometer-registry-{system} in your runtime classpath is enough for Spring Boot to configure the registry.

Most registries share common features. For instance, you can disable a particular registry even if the Micrometer registry implementation is on the classpath. For instance, to disable Datadog:

management.metrics.export.datadog.enabled=false

Spring Boot will also add any auto-configured registries to the global static composite registry on the Metrics class unless you explicitly tell it not to:

management.metrics.use-global-registry=false

You can register any number of MeterRegistryCustomizer beans to further configure the registry, such as applying common tags, before any meters are registered with the registry:

@Bean
MeterRegistryCustomizer<MeterRegistry> metricsCommonTags() {
	return registry -> registry.config().commonTags("region", "us-east-1");
}

You can apply customizations to particular registry implementations by being more specific about the generic type:

@Bean
MeterRegistryCustomizer<GraphiteMeterRegistry> graphiteMetricsNamingConvention() {
	return registry -> registry.config().namingConvention(MY_CUSTOM_CONVENTION);
}

With that setup in place you can inject MeterRegistry in your components and register metrics:

@Component
public class SampleBean {

	private final Counter counter;

	public SampleBean(MeterRegistry registry) {
		this.counter = registry.counter("received.messages");
	}

	public void handleMessage(String message) {
		this.counter.increment();
		// handle message implementation
	}

}

Spring Boot also configures built-in instrumentation (i.e. MeterBinder implementations) that you can control via configuration or dedicated annotation markers.

55.2 Supported monitoring systems

55.2.1 Atlas

By default, metrics are exported to Atlas running on your local machine. The location of the Atlas server to use can be provided using:

management.metrics.export.atlas.uri=https://atlas.example.com:7101/api/v1/publish

55.2.2 Datadog

Datadog registry pushes metrics to datadoghq periodically. To export metrics to Datadog, your API key must be provided:

management.metrics.export.datadog.api-key=YOUR_KEY

You can also change the interval at which metrics are sent to Datadog:

management.metrics.export.datadog.step=30s

55.2.3 Ganglia

By default, metrics are exported to Ganglia running on your local machine. The Ganglia server host and port to use can be provided using:

management.metrics.export.ganglia.host=ganglia.example.com
management.metrics.export.ganglia.port=9649

55.2.4 Graphite

By default, metrics are exported to Graphite running on your local machine. The Graphite server host and port to use can be provided using:

management.metrics.export.graphite.host=graphite.example.com
management.metrics.export.graphite.port=9004

Micrometer provides a default HierarchicalNameMapper that governs how a dimensional meter id is mapped to flat hierarchical names.

[Tip]Tip

To take control over this behaviour, define your GraphiteMeterRegistry and supply your own HierarchicalNameMapper. An auto-configured GraphiteConfig and Clock beans are provided unless you define your own:

@Bean
public GraphiteMeterRegistry graphiteMeterRegistry(GraphiteConfig config, Clock clock) {
	return new GraphiteMeterRegistry(config, clock, MY_HIERARCHICAL_MAPPER);
}

55.2.5 Influx

By default, metrics are exported to Influx running on your local machine. The location of the Influx server to use can be provided using:

management.metrics.export.influx.uri=https://influx.example.com:8086

55.2.6 JMX

Micrometer provides a hierarchical mapping to JMX, primarily as a cheap and portable way to view metrics locally. By default, metrics are exported to the metrics JMX domain. The domain to use can be provided using:

management.metrics.export.jmx.domain=com.example.app.metrics

Micrometer provides a default HierarchicalNameMapper that governs how a dimensional meter id is mapped to flat hierarchical names.

[Tip]Tip

To take control over this behaviour, define your JmxMeterRegistry and supply your own HierarchicalNameMapper. An auto-configured JmxConfig and Clock beans are provided unless you define your own:

@Bean
public JmxMeterRegistry jmxMeterRegistry(JmxConfig config, Clock clock) {
	return new JmxMeterRegistry(config, clock, MY_HIERARCHICAL_MAPPER);
}

55.2.7 New Relic

New Relic registry pushes metrics to New Relic periodically. To export metrics to New Relic, your API key and account id must be provided:

management.metrics.export.newrelic.api-key=YOUR_KEY
management.metrics.export.newrelic.account-id=YOUR_ACCOUNT_ID

You can also change the interval at which metrics are sent to New Relic:

management.metrics.export.newrelic.step=30s

55.2.8 Prometheus

Prometheus expects to scrape or poll individual app instances for metrics. Spring Boot provides an actuator endpoint available at /actuator/prometheus to present a Prometheus scrape with the appropriate format.

[Tip]Tip

The endpoint is not available by default and must be exposed, see exposing endpoints for more details.

Here is an example scrape_config to add to prometheus.yml:

scrape_configs:
  - job_name: 'spring'
	metrics_path: '/actuator/prometheus'
	static_configs:
	  - targets: ['HOST:PORT']

55.2.9 SignalFx

SignalFx registry pushes metrics to SignalFx periodically. To export metrics to SignalFx, your access token must be provided:

management.metrics.export.signalfx.access-token=YOUR_ACCESS_TOKEN

You can also change the interval at which metrics are sent to SignalFx:

management.metrics.export.signalfx.step=30s

55.2.10 Simple

Micrometer ships with a simple, in-memory backend that is automatically used as a fallback if no other registry is configured. This allows you to see what metrics are collected in the metrics endpoint.

The in-memory backend disables itself as soon as you’re using any of the other available backend. You can also disable it explicitly:

management.metrics.export.simple.enabled=false

55.2.11 StatsD

The StatsD registry pushes metrics over UDP to a StatsD agent eagerly. By default, metrics are exported to a StatsD agent running on your local machine. The StatsD agent host and port to use can be provided using:

management.metrics.export.statsd.host=statsd.example.com
management.metrics.export.statsd.port=9125

You can also change the StatsD line protocol to use (default to Datadog):

management.metrics.export.statsd.flavor=etsy

55.2.12 Wavefront

Wavefront registry pushes metrics to Wavefront periodically. If you are exporting metrics to Wavefront directly, your API token must be provided:

management.metrics.export.wavefront.api-token=YOUR_API_TOKEN

Alternatively, you may use a Wavefront sidecar or an internal proxy set up in your environment that forwards metrics data to the Wavefront API host:

management.metrics.export.wavefront.uri=proxy://localhost:2878
[Tip]Tip

If publishing metrics to a Wavefront proxy (as described in the documentation), the host must be in the proxy://HOST:PORT format.

You can also change the interval at which metrics are sent to Wavefront:

management.metrics.export.wavefront.step=30s

55.3 Supported Metrics

Spring Boot registers the following core metrics when applicable:

  • JVM metrics, report utilization of:

    • Various memory and buffer pools
    • Statistics related to garbage collection
    • Threads utilization
    • Number of classes loaded/unloaded
  • CPU metrics
  • File descriptor metrics
  • Logback metrics: record the number of events logged to Logback at each level
  • Uptime metrics: report a gauge for uptime and a fixed gauge representing the application’s absolute start time
  • Tomcat metrics
  • Spring Integration metrics

55.3.1 Spring MVC Metrics

Auto-configuration enables the instrumentation of requests handled by Spring MVC. When management.metrics.web.server.auto-time-requests is true, this instrumentation occurs for all requests. Alternatively, when set to false, you can enable instrumentation by adding @Timed to a request-handling method:

@RestController
@Timed 1
public class MyController {

	@GetMapping("/api/people")
	@Timed(extraTags = { "region", "us-east-1" }) 2
	@Timed(value = "all.people", longTask = true) 3
	public List<Person> listPeople() { ... }

}

1

A controller class to enable timings on every request handler in the controller.

2

A method to enable for an individual endpoint. This is not necessary if you have it on the class, but can be used to further customize the timer for this particular endpoint.

3

A method with longTask = true to enable a long task timer for the method. Long task timers require a separate metric name, and can be stacked with a short task timer.

By default, metrics are generated with the name, http.server.requests. The name can be customized by setting the management.metrics.web.server.requests-metric-name property.

By default, Spring MVC-related metrics are tagged with the following information:

  • method, the request’s method (for example, GET or POST).
  • uri, the request’s URI template prior to variable substitution, if possible (for example, /api/person/{id}).
  • status, the response’s HTTP status code (for example, 200 or 500).
  • exception, the simple class name of any exception that was thrown while handling the request.

To customize the tags, provide a @Bean that implements WebMvcTagsProvider.

55.3.2 Spring WebFlux Metrics

Auto-configuration enables the instrumentation of all requests handled by WebFlux controllers and functional handlers.

By default, metrics are generated with the name http.server.requests. You can customize the name by setting the management.metrics.web.server.requests-metric-name property.

By default, WebFlux-related metrics are tagged with the following information:

  • method, the request’s method (for example, GET or POST).
  • uri, the request’s URI template prior to variable substitution, if possible (for example, /api/person/{id}).
  • status, the response’s HTTP status code (for example, 200 or 500).
  • exception, the simple class name of any exception that was thrown while handling the request.

To customize the tags, provide a @Bean that implements WebFluxTagsProvider.

55.3.3 RestTemplate Metrics

The instrumentation of any RestTemplate created using the auto-configured RestTemplateBuilder is enabled. It is also possible to apply MetricsRestTemplateCustomizer manually.

By default, metrics are generated with the name, http.client.requests. The name can be customized by setting the management.metrics.web.client.requests-metric-name property.

By default, metrics generated by an instrumented RestTemplate are tagged with the following information:

  • method, the request’s method (for example, GET or POST).
  • uri, the request’s URI template prior to variable substitution, if possible (for example, /api/person/{id}).
  • status, the response’s HTTP status code (for example, 200 or 500).
  • clientName, the host portion of the URI.

To customize the tags, provide a @Bean that implements RestTemplateExchangeTagsProvider. There are convenience static functions in RestTemplateExchangeTags.

55.3.4 Cache Metrics

Auto-configuration enables the instrumentation of all available Caches on startup with metrics prefixed with cache. Cache instrumentation is standardized for a basic set of metrics. Additional, cache-specific metrics are also available.

The following cache libraries are supported:

  • Caffeine
  • EhCache 2
  • Hazelcast
  • Any compliant JCache (JSR-107) implementation

Metrics are tagged by the name of the cache and by the name of the CacheManager that is derived from the bean name.

[Note]Note

Only caches that are available on startup are bound to the registry. For caches created on-the-fly or programmatically after the startup phase, an explicit registration is required. A CacheMetricsRegistrar bean is made available to make that process easier.

55.3.5 DataSource Metrics

Auto-configuration enables the instrumentation of all available DataSource objects with a metric named jdbc. Data source instrumentation results in gauges representing the currently active, maximum allowed, and minimum allowed connections in the pool. Each of these gauges has a name that is prefixed by jdbc.

Metrics are also tagged by the name of the DataSource computed based on the bean name.

[Tip]Tip

By default, Spring Boot provides metadata for all supported data sources; you can add additional DataSourcePoolMetadataProvider beans if your favorite data source isn’t supported out of the box. See DataSourcePoolMetadataProvidersConfiguration for examples.

Also, Hikari-specific metrics are exposed with a hikaricp prefix. Each metric is tagged by the name of the Pool (can be controlled with spring.datasource.name).

55.3.6 RabbitMQ Metrics

Auto-configuration will enable the instrumentation of all available RabbitMQ connection factories with a metric named rabbitmq.

55.4 Registering custom metrics

To register custom metrics, inject MeterRegistry into your component, as shown in the following example:

class Dictionary {

	private final List<String> words = new CopyOnWriteArrayList<>();

	Dictionary(MeterRegistry registry) {
		registry.gaugeCollectionSize("dictionary.size", Tags.empty(), this.words);
	}

	// …

}

If you find that you repeatedly instrument a suite of metrics across components or applications, you may encapsulate this suite in a MeterBinder implementation. By default, metrics from all MeterBinder beans will be automatically bound to the Spring-managed MeterRegistry.

55.5 Customizing individual metrics

If you need to apply customizations to specific Meter instances you can use the io.micrometer.core.instrument.config.MeterFilter interface. By default, all MeterFilter beans will be automatically applied to the micrometer MeterRegistry.Config.

For example, if you want to rename the mytag.region tag to mytag.area for all meter IDs beginning with com.example, you can do the following:

@Bean
public MeterFilter renameRegionTagMeterFilter() {
	return MeterFilter.renameTag("com.example", "mytag.region", "mytag.area");
}

55.5.1 Per-meter properties

In addition to MeterFilter beans, it’s also possible to apply a limited set of customization on a per-meter basis using properties. Per-meter customizations apply to any all meter IDs that start with the given name. For example, the following will disable any meters that have an ID starting with example.remote

management.metrics.enable.example.remote=false

The following properties allow per-meter customization:

Table 55.1. Per-meter customizations

PropertyDescription

management.metrics.enable

Whether to deny meters from emitting any metrics.

management.metrics.distribution.percentiles-histogram

Whether to publish a histogram suitable for computing aggregable (across dimension) percentile approximations.

management.metrics.distribution.percentiles

Publish percentile values computed in your application

management.metrics.distribution.sla

Publish a cumulative histogram with buckets defined by your SLAs.


For more details on concepts behind percentiles-histogram, percentiles and sla refer to the "Histograms and percentiles" section of the micrometer documentation.

55.6 Metrics endpoint

Spring Boot provides a metrics endpoint that can be used diagnostically to examine the metrics collected by an application. The endpoint is not available by default and must be exposed, see exposing endpoints for more details.

Navigating to /actuator/metrics displays a list of available meter names. You can drill down to view information about a particular meter by providing its name as a selector, e.g. /actuator/metrics/jvm.memory.max.

[Tip]Tip

The name you use here should match the name used in the code, not the name after it has been naming-convention normalized for a monitoring system it is shipped to. In other words, if jvm.memory.max appears as jvm_memory_max in Prometheus because of its snake case naming convention, you should still use jvm.memory.max as the selector when inspecting the meter in the metrics endpoint.

You can also add any number of tag=KEY:VALUE query parameters to the end of the URL to dimensionally drill down on a meter, e.g. /actuator/metrics/jvm.memory.max?tag=area:nonheap.

[Tip]Tip

The reported measurements are the sum of the statistics of all meters matching the meter name and any tags that have been applied. So in the example above, the returned "Value" statistic is the sum of the maximum memory footprints of "Code Cache", "Compressed Class Space", and "Metaspace" areas of the heap. If you just wanted to see the maximum size for the "Metaspace", you could add an additional tag=id:Metaspace, i.e. /actuator/metrics/jvm.memory.max?tag=area:nonheap&tag=id:Metaspace.