Minimalistic zero-dependencies UDP/TCP statsd client for Node.js

There are times when you have to gather metrics and you want something simple without writing too much boilerplate, dats to your aid!

This client aims to have a simple statsd compliant API with some optional flavour for advanced usage, like: buffered metrics and either UDP/TCP transports!

Supports Node.js >=14.0.0, if you are a Node.js v12 user refer to [email protected].

• Installation
• Usage
  • Generic
  • Namespacing with Hostname/PID
  • TCP Client
• API
  • Client
    • new Client(options)
    • Client.close([cb])
    • Client.connect()
    • Client.counter(string[, value, sampling])
    • Client.timing(string, value[, sampling])
    • Client.gauge(string, value)
    • Client.set(string, value)
• Dats Mock
• CLI Interface
  • CLI Usage
• Benchmarks
• Powered Apps
• Support & Contribute
• License

The package is available at npm.

You can install it with npm

# lastest stable version
$ npm i -S @immobiliarelabs/dats
# latest development version
$ npm i -S @immobiliarelabs/dats@next

or yarn

# lastest stable version
$ yarn add @immobiliarelabs/dats
# latest development version
$ yarn @immobiliarelabs/dats@next

import Client from '@immobiliarelabs/dats';

const stats = new Client({
    host: 'udp://someip:someport',
    namespace: 'myGrafanaNamespace',
    // Optionally register a global handler to track errors.
    onError: (error) => {
        processError(error);
    },
});

// Send counter (myGrafanaNamespace.some.toCount)
stats.counter('some.toCount', 3);
stats.counter('some.toCount'); // defaults to 1
stats.counter('some.toCount', 1, 10); // set sampling to 10
// Send timing (myGrafanaNamespace.some.toTime)
stats.timing('some.toTime', 10);
stats.timing('some.toTime', 10, 0.1); // set sampling to 0.1

// Send gauge (myGrafanaNamespace.some.toGauge)
stats.gauge('some.toGauge', 23);

// Send set (myGrafanaNamespace.some.set)
stats.set('some.set', 765);

// Scope your stats per hostname and/or pid
import Client from '@immobiliarelabs/dats';

const stats = new Client({
    host: 'udp://someip:someport',
    namespace: 'myGrafanaNamespace.${hostname}.${pid}',
});

// Send counter (myGrafanaNamespace.myMachine.123.some.toCount)
stats.counter('some.toCount', 3);

If the hostname contains any ., the client will replace them with _.

import Client from '@immobiliarelabs/dats';

// TCP usage
const stats = new Client({
    host: 'tcp://someip:someport',
    namespace: 'myGrafanaNamespace.${hostname}.${pid}',
});

// Calling connect is required in TCP environment
await stats.connect();

// Send counter (myGrafanaNamespace.myMachine.123.some.toCount)
stats.counter('some.toCount', 3);

This module exports:

• Client

The statsd client

• options: configuration object.
  • host: statsd host (udp://{ip}:{port} or tcp://{ip}:{port}), you can use also ipv6. If you want to force udp6 usage use: udp6://{host}:{port}, when using TCP, you have to call the Client.connect method.
  • namespace: Optional. Prefix to use for the metrics. The metric will be sent as namespace. + the metric string. Optionally you can use ${hostname} and ${pid} placeholders in the namespace and have them substituted with the machine hostname and the process id.
  • bufferSize: Optional. Default is 0. Setting this value to a number greather than zero will activate buffered mode, which instead of sending metrics on each call, it will buffer them and send them when one of this conditions occurs: the buffer is full, or the bufferFlushTimeout has expired. Using this approach is more performant, but you must be careful to use a value compatible to the MTU available on your network, otherwise your packets might get dropped silently. See multi-metric-packets.
  • bufferFlushTimeout: Optional. Default is 100. Timeout in milliseconds to wait before flushing the metrics buffer.
  • debug: Optional. Default debuglog('dats'). The logger function.
  • udpDnsCache: Optional. Default true. Activate the cache DNS lookup for udp.
  • udpDnsCacheTTL: Optional. Default 120. Dns cache Time to live in seconds.
  • onError: Optional. Default (err) => void. Called when there is an error. Allows you to check also send errors.
  • customSocket: Optional. Default null. Custom socket used by the client, this is a feature for mocking we do not recommend using it in production.

close the client socket

• cb: optional. A callback function to call when the socket is closed. If no cb is provided a Promise is returned.

Returns: a Promise if no cb is passed.

connect the TCP socket. Calling this function is required only on TCP.

Returns: a Promise.

send a metric of type counter

• string: The metric string
• value: Optional. The metric value (Number). Defaults to 1.
• sampling: Optional. The metric sampling.

All sending errors are handled by the onError callback.

send a metric of type timing

• string: The metric string
• value: The metric value (Number).
• sampling: Optional. The metric sampling.

All sending errors are handled by the onError callback.

send a metric of type gauge

• string: The metric string
• value: The metric value (Number).

All sending errors are handled by the onError callback.

send a metric of type set

• string: The metric string
• value: The metric value (Number).

All sending errors are handled by the onError callback.

Dats exports his mock, you can use it as follow:

import ClientMock from '@immobiliarelabs/dats/dist/mock';

const host = new URL(`udp://127.0.0.1:8232`);
const namespace = 'ns1';
const client = new ClientMock({ host, namespace });

client.gauge('some.metric', 100);
client.set('some.metric', 100);
// metrics is an array with all metrics sent
console.log(client.metrics);
/* stdout:
    [
        'ns1.some.metric:100|g',
        'ns1.some.metric:100|s',
    ]
*/
// Check if a metric is in the metrics array
client.hasSent('ns1.some.metric:100|s'); // -> true
client.hasSent('ns1.some.metric:10|s'); // -> false
client.hasSent(/ns1\.some\.metric:\d+\|s/); // -> true
client.hasSent(/ns1\.some\.test:\d+\|s/); // -> false

// Clean the metrics array with
client.cleanMetrics();
console.log(client.metrics);
/* stdout:
    []
*/

dats is also exposed as a CLI that can both be installed as a npm global package or a precompiled binary.

The precompile binary can be found in the release section for linux, macos or window.

$ npm i -g @immobiliarelabs/dats
dats --help
# ℹ️  The following are required input flags:
#
#         --host {string} []
#         --port {string} []
#         --type {string} [Metric type can be one of: counter, timing, gauge, set]
#         --prefix {string} [Metric prefix]
#         --namespace {string} [Metric full namespace, use dots `.` to separate metrics]
#         --value {string} [Metric value]
#         --quiet {boolean} [Suppress all console output]
#         --dryRun {boolean} [Metric wont be sent, use for debug]
#
# If unsure of output run the command prepended with `DRY_RUN=1`

The automatic benchmarking for every commit can be found at the following links: next and main.

The tests were done using autocannon pointing to an HTTP node.js Server that sends at each request a count metric. With this kind of test, we evaluate how much the library influences the application performance.

Below are reported the benchmarks with the most famous node.js statsd clients:

Base is the HTTP server without metrics.

dats was created by the amazing Node.js team at ImmobiliareLabs, the Tech dept of Immobiliare.it, the #1 real estate company in Italy.

We are currently using dats in our products as well as our internal toolings.

If you are using dats in production drop us a message.

Made with ❤️ by ImmobiliareLabs & Contributors

We'd love for you to contribute to dats! If you have any questions on how to use dats, bugs and enhancement please feel free to reach out by opening a GitHub Issue.

dats is licensed under the MIT license.
See the LICENSE file for more information.