opentracing-metrics-tracer

Exports cross-process metrics via OpenTracing instrumentation to reporters: Prometheus.
It's capable to measure operation characteristics in a distributed system like microservices.

It also makes possible to reverse engineer the infrastructure topology as we know the initiators

Available Reporters

• Prometheus via prom-client

Getting started

const MetricsTracer = require('@risingstack/opentracing-metrics-tracer')
const prometheusReporter = new MetricsTracer.PrometheusReporter()
const metricsTracer = new MetricsTracer('my-service', [prometheusReporter])

// Instrument
const span = metricsTracer.startSpan('my-operation')
span.finish()
...

app.get('/metrics', (req, res) => {
  res.set('Content-Type', MetricsTracer.PrometheusReporter.Prometheus.register.contentType)
  res.end(prometheusReporter.metrics())
})

With auto instrumentation and multiple tracers

Check out: https://github.com/RisingStack/opentracing-auto

// Prometheus metrics tracer
const MetricsTracer = require('@risingstack/opentracing-metrics-tracer')
const prometheusReporter = new MetricsTracer.PrometheusReporter()
const metricsTracer = new MetricsTracer('my-service', [prometheusReporter])

// Jaeger tracer (classic distributed tracing)
const jaeger = require('jaeger-client')
const UDPSender = require('jaeger-client/dist/src/reporters/udp_sender').default
const sampler = new jaeger.RateLimitingSampler(1)
const reporter = new jaeger.RemoteReporter(new UDPSender())
const jaegerTracer = new jaeger.Tracer('my-server-pg', reporter, sampler)

// Auto instrumentation
const Instrument = require('@risingstack/opentracing-auto')
const instrument = new Instrument({
  tracers: [metricsTracer, jaegerTracer]
})

// Rest of your code
const express = require('express')
const app = express()

app.get('/metrics', (req, res) => {
  res.set('Content-Type', MetricsTracer.PrometheusReporter.Prometheus.register.contentType)
  res.end(prometheusReporter.metrics())
})

Example

See example server.

node example/server
curl http://localhost:3000
curl http://localhost:3000/metrics

API

const Tracer = require('@risingstack/opentracing-metrics-tracer')

new Tracer(serviceKey, [reporter1, reporter2, ...])

• serviceKey String, required, unique key that identifies a specific type of service (for example: my-frontend-api)
• reporters Array of reporters, optional, default: []

OpenTracing compatible tracer, for the complete API check out the official documentation.

new Tracer.PrometheusReporter([opts])

• opts Object, optional
• opts.ignoreTags Object, optional
  • Example: { ignoreTags: { [Tags.HTTP_URL]: /\/metrics$/ } } to ignore Prometheus scraper

Creates a new Prometheus reporter.

Tracer.PrometheusReporter.Prometheus

Exposed prom-client.

Reporters

Prometheus Reporter

Exposes metrics in Prometheus format via prom-client

Metrics

• operation_duration_seconds
• http_request_duration_seconds

operation_duration_seconds

Always measured.
Sample output: Two distributed services communicate over the network.

# HELP operation_duration_seconds Duration of operations in second
# TYPE operation_duration_seconds histogram
operation_duration_seconds_bucket{le="0.005",parent_service="my-parent-service",name="my-operation" 0
operation_duration_seconds_bucket{le="0.01",parent_service="my-parent-service",name="my-operation" 0
operation_duration_seconds_bucket{le="0.025",parent_service="my-parent-service",name="my-operation" 0
operation_duration_seconds_bucket{le="0.05",parent_service="my-parent-service",name="my-operation" 0
operation_duration_seconds_bucket{le="0.1",parent_service="my-parent-service",name="my-operation" 1
operation_duration_seconds_bucket{le="0.25",parent_service="my-parent-service",name="my-operation" 1
operation_duration_seconds_bucket{le="0.5",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_bucket{le="1",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_bucket{le="2.5",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_bucket{le="5",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_bucket{le="10",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_bucket{le="+Inf",parent_service="my-parent-service",name="my-operation" 2
operation_duration_seconds_sum{parent_service="my-parent-service",name="my-operation" 0.4
operation_duration_seconds_count{parent_service="my-parent-service",name="my-operation" 2

http_request_duration_seconds

Measured only when the span is tagged with SPAN_KIND_RPC_SERVER and any of HTTP_URL, HTTP_METHOD or HTTP_STATUS_CODE.
Sample output:

# HELP http_request_handler_duration_seconds Duration of HTTP requests in second
# TYPE http_request_handler_duration_seconds histogram
http_request_handler_duration_seconds_bucket{le="0.005",parent_service="my-parent-service",method="GET",code="200",name="http_request" 0
http_request_handler_duration_seconds_bucket{le="0.01",parent_service="my-parent-service",method="GET",code="200",name="http_request" 0
http_request_handler_duration_seconds_bucket{le="0.025",parent_service="my-parent-service",method="GET",code="200",name="http_request" 0
http_request_handler_duration_seconds_bucket{le="0.05",parent_service="my-parent-service",method="GET",code="200",name="http_request" 0
http_request_handler_duration_seconds_bucket{le="0.1",parent_service="my-parent-service",method="GET",code="200",name="http_request" 1
http_request_handler_duration_seconds_bucket{le="0.25",parent_service="my-parent-service",method="GET",code="200",name="http_request" 1
http_request_handler_duration_seconds_bucket{le="0.5",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_bucket{le="1",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_bucket{le="2.5",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_bucket{le="5",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_bucket{le="10",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_bucket{le="+Inf",parent_service="my-parent-service",method="GET",code="200",name="http_request" 2
http_request_handler_duration_seconds_sum{parent_service="my-parent-service",method="GET",code="200",name="http_request" 0.4
http_request_handler_duration_seconds_count{parent_service="my-parent-service",method="GET",code="200",name="http_request" 2

Future and ideas

This library is new, in the future we could measure much more useful and specific metrics with it.
Please share your ideas in a form of issues or pull-requests.