🛑 This library is DEPRECATED!

There will be no new releases of this library.

We urge all users to migrate to OpenTelemetry. Please refer to the notice in the documentation for details.

Jaeger Bindings for OpenTracing API for Node.js

This is Jaeger's client side instrumentation library for Node.js that implements Javascript OpenTracing API 1.0.

Note that this library is not designed to run in the browser, only in the Node.js-backend servers. For browser-only version, see https://github.com/jaegertracing/jaeger-client-javascript.

See the OpenTracing tutorial for the introduction on using the OpenTracing API and the Jaeger SDK.

Contributing and Developing

Please see CONTRIBUTING.md.

Installation

npm install --save jaeger-client

Initialization

The Tracer defaults to sending spans over UDP to the jaeger-agent running on localhost; the jaeger-agent handles forwarding the spans to the jaeger-collector. When you are instantiating your client instance you can specify the sampler of your choice. The library support the following samplers:

More information about sampling can be found here

var initTracer = require('jaeger-client').initTracer;

// See schema https://github.com/jaegertracing/jaeger-client-node/blob/master/src/configuration.js#L37
var config = {
  serviceName: 'my-awesome-service',
};
var options = {
  tags: {
    'my-awesome-service.version': '1.1.2',
  },
  metrics: metrics,
  logger: logger,
};
var tracer = initTracer(config, options);

Environment variables

The tracer can be initialized with values coming from environment variables:

var tracer = initTracerFromEnv(config, options);

None of the env vars are required and all of them can be overridden via properties on the config object.

By default, the client sends traces via UDP to the agent at localhost:6832. Use JAEGER_AGENT_HOST and JAEGER_AGENT_PORT to send UDP traces to a different host:port. If JAEGER_ENDPOINT is set, the client sends traces to the endpoint via HTTP, making the JAEGER_AGENT_HOST and JAEGER_AGENT_PORT unused. If JAEGER_ENDPOINT is secured, HTTP basic authentication can be performed by setting the JAEGER_USER and JAEGER_PASSWORD environment variables.

Reporting spans via HTTP

UDP has a hard size limit of 65,507 bytes; if the span is larger than this limit, the tracer will drop the span. To circumvent this, you can configure the tracer to directly send spans to the jaeger-collector over HTTP (skipping the jaeger-agent altogether).

var initTracer = require('jaeger-client').initTracer;

// See schema https://github.com/jaegertracing/jaeger-client-node/blob/master/src/configuration.js#L37
var config = {
  serviceName: 'my-awesome-service',
  reporter: {
    // Provide the traces endpoint; this forces the client to connect directly to the Collector and send
    // spans over HTTP
    collectorEndpoint: 'http://jaeger-collector:14268/api/traces',
    // Provide username and password if authentication is enabled in the Collector
    // username: '',
    // password: '',
  },
};
var options = {
  tags: {
    'my-awesome-service.version': '1.1.2',
  },
  metrics: metrics,
  logger: logger,
};
var tracer = initTracer(config, options);

Metrics and Logging

The metrics and logger objects shown in the above example must satisfy the MetricsFactory and Logger APIs respectively.

Prometheus metrics

This module brings a Prometheus(prom-client) integration to the internal Jaeger metrics. The way to initialize the tracer with Prometheus metrics:

var PrometheusMetricsFactory = require('jaeger-client').PrometheusMetricsFactory;
var promClient = require('prom-client');

var config = {
  serviceName: 'my-awesome-service',
};
var namespace = config.serviceName;
var metrics = new PrometheusMetricsFactory(promClient, namespace);
var options = {
  metrics: metrics,
};
var tracer = initTracer(config, options);

Usage

The Tracer instance created by initTracer is OpenTracing-1.0 compliant. See opentracing-javascript for usage examples. Ensure that tracer.close() is called on application exit to flush buffered traces.

TChannel Span Bridging

Because tchannel-node does not have instrumentation for OpenTracing, Jaeger-Client exposes methods wrapping tchannel handlers, and encoded channels. An encoded channel is a channel wrapped in either a thrift encoder TChannelAsThrift, or json encoder TChannelAsJson. To wrap a server handler for thrift one can initialize a tchannel bridge, and wrap the encoded handler function with a tracedHandler decorator. The tchannel bridge takes an OpenTracing tracer, and a context factory. The context factory must be a function that returns a context with the methods 'getSpan', and 'setSpan' which retrieve and assign the span to the context respectively.

import { TChannelBridge } from 'jaeger-client';
import Context from 'some-conformant-context';

function contextFactory() {
  return new Context();
}

let bridge = new TChannelBridge(tracer, { contextFactory: contextFactory });
let server = new TChannel({ serviceName: 'server' });
server.listen(4040, '127.0.0.1');
let serverThriftChannel = TChannelAsThrift({
  channel: server,
  entryPoint: path.join(__dirname, 'thrift', 'echo.thrift'), // file path to a thrift file
});

let perProcessOptions = {};
serverThriftChannel.register(
  server,
  'Echo::echo',
  perProcessOptions,
  bridge.tracedHandler((perProcessOptions, req, head, body, callback) => {
    /* Your handler code goes here. */
  })
);

Outbound calls can be made in two ways, shown below.

Using encoded channel to create a request and calling request.send()

import { TChannelBridge } from 'jaeger-client';

let bridge = new TChannelBridge(tracer);
// Create the toplevel client channel.
let client = new TChannel();

// Create the client subchannel that makes requests.
let clientSubChannel = client.makeSubChannel({
  serviceName: 'server',
  peers: ['127.0.0.1:4040'],
});

let encodedThriftChannel = TChannelAsThrift({
  channel: clientSubChannel,
  entryPoint: path.join(__dirname, 'thrift', 'echo.thrift'), // file path to a thrift file
});

// wrap encodedThriftChannel in a tracing decorator
let tracedChannel = bridge.tracedChannel(encodedThriftChannel);

// The encodedThriftChannel's (also true for json encoded channels) request object can call 'send' directly.
let req = tracedChannel.request({
  serviceName: 'server',
  context: context, // must be passed through from the service handler shown above
  headers: { cn: 'echo' },
});

// headers should contain your outgoing tchannel headers if any.
// In this instance 'send' is being called on the request object, and not the channel.
req.send('Echo::echo', headers, { value: 'some-string' });

Using top level channel to create a request and calling encodedChannel.send(request)

let tracedChannel = bridge.tracedChannel(encodedThriftChannel);

// tracedChannel.channel refers to encodedThriftChannel's inner channel which
// is clientSubChannel in this instance.
let req = tracedChannel.channel.request({
  serviceName: 'server',
  headers: { cn: 'echo' },
  context: context, // must be passed through from the service handler shown above
  timeout: someTimeout,
});
// send() can be called directly on the tracing decorator
tracedChannel.send(req, 'Echo::echo', o.headers, { value: 'some-string' }, clientCallback);

Debug Traces (Forced Sampling)

Programmatically

The OpenTracing API defines a sampling.priority standard tag that can be used to affect the sampling of a span and its children:

span.setTag(opentracing_tags.SAMPLING_PRIORITY, 1);

Via HTTP Headers

Jaeger Tracer also understands a special HTTP Header jaeger-debug-id, which can be set in the incoming request, e.g.

curl -H "jaeger-debug-id: some-correlation-id" http://myhost.com

When Jaeger sees this header in the request that otherwise has no tracing context, it ensures that the new trace started for this request will be sampled in the "debug" mode (meaning it should survive all downsampling that might happen in the collection pipeline), and the root span will have a tag as if this statement was executed:

span.setTag('jaeger-debug-id', 'some-correlation-id');

This allows using Jaeger UI to find the trace by this tag.

Trace Buffer

Specify the reporter's flush interval (ms) with config.reporter.flushIntervalMs or JAEGER_REPORTER_FLUSH_INTERVAL. The default is 1000 ms.

Calling .close() on the tracer will properly flush and close composed objects, including the reporter and sampler. This prevents dropped traces in the event of an error or unexpected early termination prior to normal periodic flushing.

tracer.close(cb?)

Zipkin Compatibility

Support for Zipkin's B3 Propagation HTTP headers is provided by the ZipkinB3TextMapCodec, which can be configured instead of the default TextMapCodec.

The new codec can be used by registering it with a tracer instance as both an injector and an extractor:

let codec = new ZipkinB3TextMapCodec({ urlEncoding: true });

tracer.registerInjector(opentracing.FORMAT_HTTP_HEADERS, codec);
tracer.registerExtractor(opentracing.FORMAT_HTTP_HEADERS, codec);

This can prove useful when compatibility with existing Zipkin tracing/instrumentation is desired.

Webpack Compatibility

In order to bundle the library using webpack, e.g. for uploading code to an AWS Lambda function, it is required to copy the Jaeger thrift definition file into the output directory of the bundle:

{
  plugins: [
    new CopyPlugin([
      {
        from: require.resolve('jaeger-client/dist/src/jaeger-idl/thrift/jaeger.thrift'),
        to: 'jaeger-idl/thrift/jaeger.thrift',
      },
      {
        from: require.resolve('jaeger-client/dist/src/thriftrw-idl/agent.thrift'),
        to: 'thriftrw-idl/agent.thrift',
      },
    ]),
  ];
}

License

Apache 2.0 License.