SMetrics

Features:

• tagless final api via cats
• improved abstraction for matching labels names and label values
• resource safety
• detaching metrics backend - prometheus dependency is one of the backend implementations

Example of http metrics

let's declare interface for metrics, that seems reasonable for http client

trait HttpMetrics[F[_]] {

  def latency(client: String, method: String, resource: String, duration: FiniteDuration): F[Unit]

  def count(client: String, method: String, resource: String, success: Boolean): F[Unit]

  def enqueue(client: String): F[Unit]

  def dequeue(client: String): F[Unit]
}

now let's implement HttpMetrics using CollectorRegistry

  import com.evolutiongaming.smetrics.MetricsHelper.*
  import com.evolutiongaming.smetrics.{CollectorRegistry, LabelNames, Quantile, Quantiles}

  def httpMetrics[F[_]: Monad](collectorRegistry: CollectorRegistry[F]): Resource[F, HttpMetrics[F]] = {
    for {
      latencySummary <- collectorRegistry.summary(
        "http_client_latency",
        "Latency of HTTP requests processing in seconds",
        Quantiles(
          Quantile(value = 0.5, error = 0.05),
          Quantile(value = 0.9, error = 0.05),
          Quantile(value = 0.99, error = 0.005)),
        LabelNames("client_name", "method", "resource"))
      
      resultCounter <- collectorRegistry.counter(
        "http_client_response_result",
        "Status of HTTP responses",
        LabelNames("client_name", "method", "resource", "result"))
      
      queueGauge <- collectorRegistry.gauge(
        "http_client_queue",
        "Queue of incoming http calls",
        LabelNames("client_name"))
    } yield {
      new HttpMetrics[F] {
    
        def latency(client: String, method: String, resource: String, duration: FiniteDuration) = {
          latencySummary
            .labels(client, method, resource)
            .observe(duration.toNanos.nanosToSeconds)
        }
    
        def count(client: String, method: String, resource: String, success: Boolean) = {
          resultCounter
            .labels(client, method, resource, if (success) "success" else "error")
            .inc()
        }
    
        def enqueue(client: String) = {
          queueGauge
            .labels(client)
            .inc()
        }
    
        def dequeue(client: String) = {
          queueGauge
            .labels(client)
            .dec()
        }
      }
    }
  }

So you can see, httpMetrics function requires instance of CollectorRegistry. Right now there are two options available:

• CollectorRegistry.empty - often used in tests
• CollectorRegistryPrometheus - which will create smetrics.CollectorRegistry out of io.prometheus.CollectorRegistry using prometheus/client_java under the hood

Return type is Resource[F, _] rather than F[_] because most often underlying metrics implementation upon a call registers your metrics with shared registry. Hence release hook of Resource being used in order to de-register particular metrics from shared registry.

Setup

addSbtPlugin("com.evolution" % "sbt-artifactory-plugin" % "0.0.2")

libraryDependencies += "com.evolutiongaming" %% "smetrics" % "0.3.1"

libraryDependencies += "com.evolutiongaming" %% "smetrics-prometheus" % "0.3.1"

For prometheus version 1.x.x+ use prometheus_v1 module:

libraryDependencies += "com.evolutiongaming" %% "smetrics-prometheus-v1" % "0.3.1"

For sttp3 use smetrics-sttp3 module:

libraryDependencies += "com.evolutiongaming" %% "smetrics-sttp3" % "x.y.z"