Bundle-time macros with Bun

#​641 — June 1, 2023

Read on the Web

JavaScript Weekly

JavaScript Macros in Bun — Not content with giving the JavaScript world a brand new bundler, Bun’s Jarred has taken it a bit further: ‘macros’ that run at bundle time with the result being directly inlined into your code. They use stage 3 annotated import statements (so may become regular JS eventually) and Jarred shows off some use cases here.

Jarred Sumner

▶  Discussing the Future of React with Two Core Team Members — As part of the celebrations around React’s 10th birthday, Vercel’s Delba de Oliveira interviewed Andrew Clark and Sebastian Markbåge of the React core team on modern topics including server components, Suspense, Actions, and the next steps React will take.

Delba de Oliveira (Vercel)

URL Validation 101 — with Snyk — Keep script injection and server side request forgery out of your JavaScript applications with Snyk.

Snyk sponsor

Aimless.js: The ‘Missing’ JavaScript Randomness Library — If you’ve been itching for functions to give you random characters, numbers from custom distributions, random sequences, random items, weighted random numbers, or more, this is for you.

Christopher Cavalea

⚡️ IN BRIEF:

???? Two big birthdays in the past week: Node.js turned 14 and React turned 10.

Intel and Google have been working together on the Compute Pressure API which you can use, in origin trial, in Chrome 115. It provides a way to measure the ‘computational strain’ of the system your code is running on to then adjust how your app operates accordingly.

???? Misko Hevery, creator of Angular and now Qwik, was on the Stack Overflow Podcast talking about ▶️ how he’s ‘dehydrating the Web’ in an ongoing quest for higher performance.

‘Deferring Module Evaluation’ is a TC39 proposal for basically being able to lazily load modules which are executed only upon use.

You can now use Node.js built-in modules on Deno Deploy, making it easier to run existing JS apps ‘at the edge’ on their platform.

Tom Preston-Werner, GitHub co-founder and Redwood co-creator, writes about how Redwood is going ‘all in on React Server Components’ and announces the first Redwood conference this September.

SupportsCSS: Feature Detection for Modern CSS — Inspired by Modernizr, this script expands on the capabilities of CSS’s @supports feature by adding classes to your HTML and exposing a results object so you can run custom tests live in the browser.

Stephanie Eckles

????  JSDayIE 2023: The First JavaScript Conference in Ireland Is Back! — Join us on September 26th in Dublin to experience everything the Irish JavaScript community and Ireland have to offer.

JSDayIE sponsor

RELEASES:

Electron 25 – Now offering a Chrome-powered net.fetch (as opposed to Node’s flavor of fetch()).

Neutralinojs 4.12 – Desktop app dev framework.

Parcel 2.9

Bootstrap 5.3.0

node-oracledb 6.0

???? Articles & Tutorials

How to Draw Any Regular Shape with Just One JavaScript Function — Mozilla/MDN has a (relatively) new blog — here they share how to use JavaScript to draw any regular shape to a HTML canvas with a single function, and how to modify it to draw multiple shapes.

Ruth John

Sharing WebSocket Connections between Tabs and Windows — If your users open your app in multiple tabs or windows, being able to share a WebSocket connection could be a lot more efficient both at the client and server end. Is it possible? With SharedWorkers (supported by all major browsers, except Chrome on Android), you can.

Szymon Chmal

▶  How React Works Under The Hood: 2023 Edition — Quite a lot of depth for a mere 13-minutes. Covers React broadly, along with the pros and cons of JSX, how the virtual DOM approach works, plus React’s diffing algorithm.

FrontStart

Connected APM and Infrastructure Monitoring at Up to 1/3rd the Cost of Datadog

New Relic sponsor

Why Does a Farmer Emoji Have a Length of 7 in JS? — This is a nicely accessible introduction to a topic that often causes confusion. Get ready to learn a little about grapheme clusters, scalars, and code units.

Evan Hahn

The Evolution of React APIs and Code Reuse — An exploration of common React patterns, from mixins and higher-order components to render props & hooks.

Rem / Frontend Mastery

▶  ‘I Tried 8 Different Postgres ORMs’ — A modern, fast-paced, and slightly irreverent look at working with Postgres from back-end JavaScript runtimes. (9 minutes.)

Beyond Fireship

Automatically Unsubscribe from RxJS Observables in Angular Components

Rob Bell

Deploying Vanilla Node.js Apps on Fly.io

Sam Ruby (Fly)

Asynchronous JavaScript for Beginners

Daniel Agantem

???? Code & Tools

Svelvet: A Svelte Component Library for Interactive Node-Based Diagrams — You can create flow diagrams using pre-built components with seamless zooming and panning, draggable interactivity, customizable edges/nodes, etc.

Svelvet Team

Inkline 4.0: A Customizable Vue.js 3 UI/UX Library — A design system and numerous customizable components designed for mobile-first (but desktop friendly) and built with accessibility in mind.

Alex Grozav

Handsontable: Data Editor for React — There’s a lot of JS data grids out there, but none quite like Handsontable: Excel-like design, keyboard shortcuts, and React support make it the perfect data editor for web apps.

Handsontable Data Grid sponsor

JECS: An Entity Component System (ECS) for JSEntity component systems are common in game development as they provide a lot of flexibility for managing the many objects present in games.

Stuhl

Color Names: A Gigantic List of Handpicked Color Names — At first, I thought this was just going to be a hundred or so obvious choices like the named colors in HTML (???? rebeccapurple) but it’s actually over 30,000 colors with a fantastic array of names and uses, plus JavaScript functions to work with them.

David Aerne

fastgron: High-Performance JSON to GRON Converter‘gron’ is a transpiled form of JSON into individual assignment operations which makes it easier to grep. (fastgron is itself written in C++.)

Adam Ritter

Orama 1.0.3
↳ In-memory, typo-tolerant, text search engine.

Perspective 2.2
↳ WASM-powered high-perf data visualization component.

Javet 2.2 – Java + V8. Embed JS into Java.

OverlayScrollbars 2.2 – JS custom scrollbar plugin.

html-react-parser 4.0 – HTML to React parser.

Vuetify 3.3.2 – Vue component framework.

React Slider 10.2Demos.

???? Jobs

Find JavaScript Jobs with Hired — Hired makes job hunting easy-instead of chasing recruiters, companies approach you with salary details up front. Create a free profile now.

Hired

????‍???? Got a job listing to share? Here’s how.

????  Live-streaming from JavaScript

JSFiddle: Did You Know You Can Stream on Twitch from JavaScript? — This is an interesting experiment, but I tried it and it worked! Twitch added support for WebRTC ingestion a while ago and this CodePen example ties that together with fewer than 50 lines of JavaScript to send your preferred video source to the popular livestreaming service. It’s hardly OBS, but I imagine someone will rustle up something fancier with this approach in time.

Sean DuBois on JSFiddle

Measure Node.js server response time with N|Solid

As software developers, we constantly face new challenges in an ever-changing ecosystem. However, we must always remember the importance of addressing performance and security concerns, which remain at the top of our priority list.

To ensure that our applications based on Node.js can meet our performance and scalability needs without compromising security or incurring costly infrastructure changes, we must be aware of the importance of network optimization in Node.js.

The Impact of Latency/Ping Time on the Performance and Speed of Your Node.js Application

IMG – Ping Cats – via GIPHY

This communication, known as network ping time or latency, is a crucial factor that impacts the performance and speed of your application. Knowing how to measure network ping time between the browser and the server is essential for developers who want to optimize their applications and provide a better user experience. _Have you ever wondered how long it takes for your application to communicate with the server? _

Network Optimization in Node.js

To ensure the optimal performance and scalability of our Node.js applications, we must accurately measure our HTTP server’s connection and response time. Doing so enables us to identify and address potential bottlenecks without compromising security or incurring unnecessary infrastructure changes.

Before delving deeper into measuring connection and response time, let’s explore fundamental concepts and critical differentiators in the network landscape.

HTTP vs. WebSocket:

HTTP and WebSocket are communication protocols used in web development but serve different purposes. HTTP is a stateless protocol commonly used for client-server communication, while WebSocket enables full-duplex communication between clients and servers, allowing real-time data exchange.

Types of Connections and Versions:

When creating APIs, HTTP as a protocol and standard has different versions, such as HTTP 1.1 and 2.0. Additionally, APIs may use alternative protocols like gRPC, which offer different features and capabilities. Understanding these options empowers developers to choose the most suitable tools for their web servers.

TCP/IP Basics:

The Transmission Control Protocol (TCP) and Internet Protocol (IP) are fundamental protocols that form the backbone of computer networks. Among TCP’s critical processes is the three-way handshake, which plays a vital role in establishing a secure and dependable connection between two endpoints. This handshake ensures the orderly and reliable transmission of data. TLS/SSL encryption enhances security, adding an extra layer of protection to the communication between the client and the server.

HTTP vs. HTTPS:

HTTP operates over plain text, which exposes the data being transmitted to potential eavesdropping and tampering.
HTTPS, on the other hand, secures communication through the use of SSL/TLS encryption, providing confidentiality and integrity.
Understanding the trade-offs between HTTP and HTTPS is crucial to making informed data security decisions.

Building a Solid Foundation: Understanding the Three-Way Handshake for Reliable Connections

To evaluate the performance of our HTTP server, we need to differentiate between connection latency and server response time. Connection latency refers to the time it takes for the initial three-way handshake process to complete before data transmission can occur. On the other hand, server response time measures the duration from when the server receives a request to when it generates and sends the response back to the client.

The three-way handshake is a fundamental process in establishing a TCP (Transmission Control Protocol) connection between a client and a server in a network. It involves three steps, a “three-way handshake.” This handshake establishes a reliable and ordered communication channel between the two endpoints.

Here’s a breakdown of the three steps involved in the three-way handshake:

__SYN (Synchronize)__: The client initiates the connection by sending an SYN packet (synchronize) to the server. This packet contains a randomly generated sequence number to initiate the communication.
__SYN-ACK (Synchronize-Acknowledge)__: Upon receiving the SYN packet, the server acknowledges the request by sending an SYN-ACK packet back to the client. The SYN-ACK packet includes its own randomly generated sequence number and an acknowledgment number equal to the client’s sequence number plus one.
__ACK (Acknowledge)__: Finally, the client sends an ACK packet (acknowledge) to the server, confirming the receipt of the SYN-ACK packet. This packet also contains the acknowledgment number equal to the server’s sequence plus one.

Once this three-way handshake process is completed, the client and the server have agreed upon initial sequence numbers, and a reliable connection is established between them. This connection allows for data transmission with proper sequencing and error detection mechanisms, ensuring that the information sent between the client and server is reliable and accurate.

The three-way handshake is essential to establishing TCP connections and is performed before any data transmission can occur. It plays a critical role in ensuring the integrity and reliability of the communication channel, providing a solid foundation for subsequent data exchange between the client and server.

Create a self-serve diagnostic tool for a server-rendered page in Node.js.

The idea is to share an easy-to-follow recipe that will help you create your tool, so let’s start with the ingredients and end with the steps to create a self-serve diagnostic tool for a server-rendered page in Node.js.

Ingredients:

Node.js & NPM installation – https://nodejs.org/

Fastify.js – https://www.fastify.io/

Instructions:

1. Setup a Node.js Project
Use NPM to create your Node project:

$ mkdir diagnostic-tool-nodejs
$ cd diagnostic-tool-nodejs
$ npm init -y

2. Install your NPM packages.
We have Fastify in our recipe, so we must install them first:

$ npm i fastify

3. Create the index.mjs
Create an index.mjs file in the project’s root directory and paste this fastify HTTP server sample code.

import Fastify from “fastify”;

const fastify = Fastify({
logger: true,
});

// Randomly create a timer from 100ms up to X seconds
function timer(time) {
return new Promise((resolve, reject) => {
const ms = Math.floor(Math.random() * time) + 100;
setTimeout(() => {
resolve(ms);
}, ms);
});
};

// Declare the root route and delay the response randomly
fastify.get(“/”, async function (request, reply) {
const wait = await timer(5000);
return { delayTime: wait };
});

// Run the server!
fastify.listen({ port: 3000 }, function (err, address) {
if (err) {
fastify.log.error(err);
process.exit(1);
}
});

This will start the server on port 3000, which you can access by going to http://localhost:3000 in your web browser.

Integrate with N|Solid Console

Be sure you already have N|Solid installed and running on your environment; otherwise, go to https://downloads.nodesource.com and get the installer.

Also, run the console using docker as an alternative to the local installation.

docker run -d -p 6753:6753 -p 9001:9001 -p 9002:9002 -p 9003:9003 nodesource/nsolid-console:hydrogen-alpine-latest

With the application already initialized with npm, Fastify installed, and our index.js in place, we can connect our process with N|Solid

Run the HTTP server with the NSOLID RUNTIME following the instructions on the principal console page.

IMG – Connect N|Solid

In this case, we ran the process by passing the config via environment variables and running a local installation of the Nsolid console.

NSOLID_APPNAME=”NSOLID_RESPONSE_TIME_APP” NSOLID_COMMAND=”127.0.0.1:9001″ nsolid index.mjs

If you instead use our SaaS console, you need to use the NSOLID_SAAS env instead of __NSOLID_COMMAND__.

NSOLID_APPNAME=”NSOLID_RESPONSE_TIME_APP” NSOLID_COMMAND=”XYZ.prod.proxy.saas.nodesource.io:9001″ nsolid index.mjs

After completing those steps, you should be able to watch the app and process connected to the console.

IMG – Connect N|Solid Process

GIF 1 – Connect N|Solid Process

Go to the application process and add the HTTP(S) Server 99th Percentile Duration metric to see in near-real time the HTTP server latency response time and also we have the HTTP(S) Request Median Duration.

GIF 2 – Monitor Process Metrics

After this, we should be able to generate some traffic and see how the response times behave with the sample code provided, generating some response time randomness from 100ms up to 5 secs.

To generate the traffic, we can use autocannon

npx autocannon -d 120 -R 60 localhost:3000

After running autocannon for some minutes, we can see the P99 metric of the HTTP Server. The median and compare them.

IMG – http-latency-response-time-metrics

IMG – http-request-median-duration

IMG – p99-metric

To fully utilize the metrics provided by N|Solid, it is crucial to have a comprehensive understanding of their significance. Two critical metrics offered by N|Solid are the 99th Percentile and the HTTP Median metric. These metrics play a vital role in assessing the performance of Node.js applications in production environments. By getting deeper into their practical application and importance, we can unlock the actual value of these metrics in N|Solid and make informed decisions to optimize our production systems. Let’s explore this further.

The 99th Percentile metric

The 99th percentile is a statistical measure commonly used to analyze and understand response time or latency in a system.

Imagine you have a web application that handles incoming requests. To understand how fast the server responds, you measure the time it takes for each request and gather that data. You can find the 99th percentile response time by looking at the data.

For example, __the 99th percentile response time is 500 milliseconds__.
This means that only 1% of the requests took longer than 500 milliseconds to get a response. In simpler terms, 99% of the requests were handled in 500 milliseconds or less, which is fast.

It helps you identify and address any outliers or performance bottlenecks affecting a small fraction of requests but can significantly impact the user experience or system stability. Monitoring the 99th percentile response time helps you spot any slow requests or performance issues that might affect a few users but still need attention. but can have a significant impact on user experience or system stability.

The HTTP median metric

When sorted in ascending or descending order, the median represents a dataset’s middle value.

To illustrate the difference between the 99th percentile and the median, let’s consider an example. Suppose you have a dataset of response times for a web application consisting of 10 values:
[100ms, 150ms, 200ms, 250ms, __500ms__, 600ms, 700ms, 800ms, 900ms, 1000ms].

The median response time would be the middle value when the dataset is sorted, which is the 5th value, 500ms. This means that 50% of the requests had a response time faster than 500ms, and the other 50% had a response time slower than 500ms.

Connect with NodeSource

If you have any questions, please contact us at [email protected] or through this form.

Experience the Benefits of N|Solid’s Integrated Features
Sign up for a Free Trial Today

To get the best out of Node.js and experience the benefits of its integrated features, including OpenTelemetry support, SBOM integration, and machine learning capabilities. Sign up for a free trial and see how N|Solid can help you achieve your development and operations goals. #KnowyourNode

AIOps Observability: Going Beyond Traditional APM

AIOps is an emerging technology that applies machine learning and analytics techniques to IT operations. AIOps enables IT teams to leverage advanced algorithms to identify performance issues, predict outages, and optimize system performance. Nodesource sees significant advantages for developers and teams to increase software quality by leveraging AIOPS. We have extended our platform’s (N|Solid) observability capabilities to include AIOps, enabling developers to leverage advanced machine learning and analytics techniques to optimize their Node.js applications.

Our N|Solid platform provides the most advanced visibility into Node.js applications, enabling developers to quickly identify performance issues, detect security vulnerabilities, and troubleshoot errors. N|Solid achieves this level of observability through real-time performance monitoring, comprehensive metrics, and detailed instrumentation of Node.js applications.

Last year, we integrated OpenTelemetry into our runtime and were nearing the release of an extension of this layer into our console. This advancement will further extend our platform to support AIOps. Santiago Gimeno, a Senior Architect, sums up our vision of the integration with OTel (Open Telemetry) and N|Solid:

“In today’s world, where applications are becoming more complex and distributed, having a good observability solution is more important than ever. The emergence of OpenTelemetry as the de-facto standard for observability is key. It allows application developers to select solutions that adapt better to their needs. Even more, it allows for healthy competition between observability solution vendors. We support this approach and continue to take steps to ensure N|Solid stays compliant with the OpenTelemetry specification, so everyone can use what we believe is the best observability solution for Node.js.”

Key differences between an APM and Observability

APM (Application Performance Management) and observability are both methods of monitoring and managing the performance and health of software applications, but there are key differences between the two:

Scope: APM is focused on monitoring the performance of applications, while observability is a more comprehensive approach that includes monitoring the infrastructure and application stack, as well as the performance of individual services.

Metrics: APM typically relies on predefined metrics and thresholds to identify performance issues, while observability takes a more flexible approach to collect a wide range of data, including logs, metrics, and traces.

Root cause analysis: APM is designed to quickly identify the root cause of performance issues, often through alerting and automated remediation, while observability takes a more holistic approach that emphasizes the need to understand the relationships between different parts of the system to identify and fix issues.

Proactivity: APM is often reactive, focusing on identifying and fixing issues as they arise, while observability is more proactive, focusing on continuous monitoring and analysis to identify potential issues before they become critical.

Tooling: APM is often built around specific tools and technologies designed to monitor and analyze application performance, while observability is more flexible and adaptable, focusing on integrating a wide range of tools and technologies to provide a comprehensive system view.

You need to have both; they are important to monitoring and managing the performance and health of software applications.

This is why N|Solid is not only an APM but also has observability within its functionalities. And now, with the implementation of ML and SBOM, it goes beyond APM and supports the growing discipline of AIOps.

AIOps: Fundamental concept in Modern IT Operations

AIOps (Artificial Intelligence for IT operations) is an approach to IT operations that leverages machine learning and artificial intelligence to automate and optimize IT tasks. AIOps aims to enhance the efficiency and effectiveness of IT operations by leveraging the vast amount of data generated by various IT systems and applications.

Observability refers to IT teams’ ability to observe and understand the behavior of complex systems in real-time, using a combination of monitoring, logging, and analytics tools.

AIOps and observability enable IT teams to proactively monitor and manage IT systems, applications, and infrastructure, allowing them to identify and resolve issues quickly. AIops uses machine learning and AI algorithms to identify patterns in large amounts of data, while observability provides the visibility and context needed to understand the behavior of complex systems.

Modern Observability in Place

Support for open-source tracing tools and standards like OpenTelemetry facilitates team collaboration in resolving issues. Open Telemetry is the second most active CNCF project, behind only Kubernetes, showcasing its importance to the industry.

_Image Twitter Michael Haberman @hab_mic
_

Following this standard, N|Solid (since N|Solid 4.8.0) supports OTEL:
– Implements the OpenTelemetry TraceAPI, allowing users to use the de-facto standard API to instrument their code.
– It supports using many instrumentation modules available in the Open Telemetry ecosystem. It supports exporting traces using the Open Telemetry Protocol(OTLP) over HTTP.
– With this feature is now possible to send N|Solid runtime monitoring information (metrics and traces) to backends supporting the Open Telemetry standard like multiple APMS (Dynatrace, Datadog, Newrelic, etc.).

Additionally, we included OTEL in the ‘APM performance dashboard,’ an open-source tool we released to the community, enabling developers and organizations to understand the impact of APM tools’ performance.

_Img APM’s Performance Dashboard View
_

Recent enhancements to the tool include the following:

Updated the data with N|Solid 4.8.0 -> 16.16.0 and 14.20.0
Added a few new tests: especially with different solutions for graphql.
Added more APMs: opentelemetry, AppDynamics

Added testing of N|Solid against Datadog, Dynatrace, and NewRelic.

Do you want to implement OTel in your Node.js application? ????

Enriching telemetry data with metadata is an important aspect of observability, and OpenTelemetry provides a flexible and extensible framework for doing so. However, there can be challenges in implementing this in practice, especially when dealing with multiple tools and technologies.

One approach to addressing this challenge is to use a centralized configuration management tool to ensure consistency in metadata enrichment across your observability stack. Please review the following articles to give you an accurate guide to implementing Opentelemetry in your project.

Enhance Observability with Opentelemetry tracing – Part 1

Instrument your Nodejs Applications with Open Source Tools – Part 2

However, if you want this implementation out of the box and have other useful features, we invite you to try N|Solid.

Conclusion

N|Solid Supports OpenTelemetry Features, Integrates SBOM and ML at its Core.

By supporting OpenTelemetry features, N|Solid provides seamless integration with this framework, enabling customers to understand their applications, infrastructure behavior, and performance. This integration enhances the ability of developers and operators to troubleshoot issues, identify bottlenecks, and optimize application performance.

N|Solid’s integration of Software Bill of Materials (SBOM) provides a comprehensive list of all software components used in an application, including open-source libraries and dependencies, which helps organizations to mitigate security risks and ensure compliance with regulations. By integrating SBOM at its core, N|Solid provides organizations with an efficient and reliable way to manage the security and compliance of their software applications.

Finally, N|Solid’s integration of machine learning (ML) at its core is another critical feature that helps to identify patterns and anomalies in data, allowing developers and operators to gain insights that are not easily detectable using traditional monitoring tools. This integration of ML at the core of N|Solid enables organizations to improve the overall reliability, performance, and security of their applications and services.

N|Solid’s support of OpenTelemetry features, integration of SBOM, and integration of ML at its core provides developers and operators with a comprehensive set of tools to manage and optimize their applications and infrastructure, making N|Solid a valuable platform in the modern software development and operations landscape.

Ready to connect?

If you want to know more about our APM’S Benchmark project and get the most out of your Node.js application, read this incredible article by our VP of engineering, Adrián Estrada, ‘In-depth analysis of the APMs performance cost in Node.js.

We also invite you to ????️ Use the ✨APM’s Performance Dashboard✨here:
???? Read the full blog post here: https://nsrc.io/4xFaster
???? Contribute here: https://github.com/nodesource/node-APMs-benchmark
If you have any questions, please contact us at [email protected] or through this form.

Experience the Benefits of N|Solid’s Integrated Features
Sign up for a Free Trial Today

To get the best out of Node.js and experience the benefits of its integrated features, including OpenTelemetry support, SBOM integration, and machine learning capabilities. Sign up for a free trial and see how N|Solid can help you achieve your development and operations goals. #KnowyourNode

Instrument your Nodejs Applications with Open Source Tools – Part 2

As we mentioned in the previous article, at NodeSource, we are dedicated to observability in our day-to-day, and we know that a great way to extend our reach and interoperability is to include the Opentelemetry framework as a standard in our development flows; because in the end our vision is to achieve high-performance software, and it is what we want to accompany the journey of developers in their Node.js base applications.

With this, we know that understanding the bases was very important to know the standard and its scope, but that it is necessary to put it into practice. How to integrate Opentelemetry in our application?; and although NodeSource has direct integration into its product in addition to more than 10 key functionalities in N|Solid, that extend the offer of a traditional APM, as you know, we are great contributors to the Open Source project, we also support the binary distributions of the Node.js project, our DNA is always helping the community and showing you how through Open Source tools you can still increase the visibility. So through this article, we want to share how to set up OpenTelemetry with Open Source tools.

In this article, you will find __How to Apply the OpenTelemetry OS framework in your Node.js Application__, which includes:

Step 1: Export data to the backend

Step 2: Set up the Open Telemetry SDK
__Step 3__: Inspect Prometheus to review we’re receiving data

Step 4: Inspect Jaeger to review we’re receiving data

Step 5: Getting deeper at Jaeger 👀

Note: This article is an extension of our talk at NodeConf.EU, where we had the opportunity to share the talk:

__Dot, line, Plane Trace!__
__Instrument your Node.js applications with Open Source Software__
Get insights into the current state of your running applications/services through OpenTelemetry. It has never been as easy as now to collect data with Open Source SDKs and tools that will help you extract metrics, generate logs and traces and export this data in a standardized format to be analyzed using the best practices. In this talk, We’ll show how easy it is to integrate OpenTelemetry in your Node.js applications and how to get the most out of it using Open Source tools.

To see the talks from this incredible conference, you can watch all sessions through live-stream links below 👇
– Day 1️⃣ – https://youtu.be/1WvHT7FgrAo
– Day 2️⃣ – https://youtu.be/R2RMGQhWyCk
– Day 3️⃣ – https://youtu.be/enklsLqkVdk

Now we are ready to start 💪 📖 👇

Apply the OpenTelemetry OS framework in your Node.js Application

So, going back to the distributed example we described in our previous article, here we can see what the architecture looks like this after adding observability.

Every service will collect signals by using the OpenTelemetry Node.js SDK and export the data to specific backends so we can analyze it.

We are going to use the following:

JAEGER for Traces and Logs.

Prometheus to visualize the metrics.

_Note: _Jaeger and Prometheus are probably the most popular open-source tools in space.

Step 1: Export data to the backend

How the data is exported to the backends differs:
To send data to _JAEGER__, we will use OTLP over HTTP, whereas for _Prometheus__, the data will be pulled from the services using HTTP.

First, we will show you how easy it is to set up the OpenTelemetry SDK to add observability to our applications.

### Step 2: Set up the OpenTelemetry SDK

First, we have the providers in charge of collecting the signals, in our case __NodeTracerProvider__ for traces and __MeterProvider__ for metrics.
Then the exporters send the collected data to the specific backends.
The Resource contains attributes describing the current process, in our case, __ServiceName__ and __Container. Id’s__. The name of these attributes is well defined by the spec (it’s in the __semantic_conventions module__) and will allow us to differentiate where a specific signal comes from.

So to set up traces and metrics, the process is basically the same: we create the provider passing the Resource, then register the specific exporter.

We also register instrumentations of specific modules (either core modules or popular userspace modules), which provide automatic Span creation of those modules.

Finally, the only important thing to remember is that we need to initialize OpenTelemetry before our actual code; the reason is these instrumentation modules (in our case for __http__ and fastify) __monkeypatch__ the module they’re instrumenting.

Also, we create the __meter instruments__ because we will use them on every service: an __HTTP request counter__ and a couple of observable gauges for __CPU usage__ and __ELU usage__.

So let’s spin the application now and send a request to the API. It returns a 401 Not Authorized. Before trying to figure out what’s going on, let’s see if Prometheus and jaeger are actually receiving data.

Step 3: Inspect Prometheus to review we’re receiving data

Let’s look at Prometheus first:
Looking at the HTTP requests counter, we can see there are 2 data points: one for the __API service__ and another one for the __AUTH service__. Notice that the data we had in the Resource is __service_name__ and __container_id__. We also can see the process_cpu is collecting data for the 4 services. The same is true for __thread_elu__.

Step 4: Inspect Jaeger to review we’re receiving data

Let’s look at Jaeger now:
We can see that one trace corresponding to the __HTTP request__ has been generated.

Also, look at this chart where the points represent traces, the X-axis is the timestamp, and the Y-axis is the duration. If we inspect the trace, we can see it consists of 3 spans, where every span represents an __HTTP transaction__, and it has been automatically generated by the instrumentation-HTTP modules:

The 1st span is an HTTP server transaction in the API service (the incoming HTTP request).
The 2nd span represents a POST request to AUTH from API.
The 3rd one represents the incoming HTTP POST in AUTH. If we inspect a bit this last span, apart from the typical attributes associated with the request (HTTP method, request_url, status_code…).

We can see there’s a Log associated with the Span this makes it very useful as we can know exactly which request caused the error. By inspecting it, we found out that the reason for the failure was missing the auth token.

This piece of information wasn’t generated automatically, though, but it’s very easy to do. So in the verify route from the service, in case there’s an error verifying the token, we retrieve the active span from the current context and just call __recordException()__ with the error. As simple as that.

Well, so far, so good. Knowing what the problem is, let’s add the auth token and check if everything works:

curl http://localhost:9000/ -H “Authorization: Bearer eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiIiLCJpYXQiOjE2NjIxMTQyMjAsImV4cCI6MTY5MzY1MDIyMCwiYXVkIjoid3d3LmV4YW1wbGUuY29tIiwic3ViIjoiIiwibGljZW5zZUtleSI6ImZmZmZmLWZmZmZmLWZmZmZmLWZmZmZmLWZmZmZmIiwiZW1haWwiOiJqcm9ja2V0QGV4YW1wbGUuY29tIn0.PYQoR-62ba9R6HCxxumajVWZYyvUWNnFSUEoJBj5t9I”

Ok, now it succeeded. Let’s look at Jaeger now. We can see the new trace here, and we can see that it contains 7 spans, and no error was generated.

Now, it’s time to show one very nice feature of Jaeger. We can compare both traces, and we can see in grey the Spans that are equal, whereas we can see in Green the Spans that are new. So just by looking at this overview, we can see that if we’re correctly Authorized, the API sends a GET request to SERVICE1, which then performs a couple of operations against POSTGRES. If we inspect one of the POSTGRES spans (the query), we can see useful information there, such as the actual QUERY. This is possible because we have registered the instrumentation-pg module in SERVICE1.

And finally, let’s do a more interesting experiment. We will inject load to the application for 20 seconds with autocannon…

If we look at the latency chart, we see some interesting data: up until at least the 90th percentile, the latency is basically below 300ms, whereas starting at least from 97.5%, the latency goes up a lot. More than 3secs. This is Unacceptable 🧐. Let’s see if we can figure out what’s going on 💪.

Step 5: Getting deeper at Jaeger 👀

Looking at Jaeger and limiting this to like 500 spans, we can see that the graph here depicts what the latency char showed. Most of the requests are fast, whereas there are some significant outliers.

Let’s compare one of the fast vs. slow traces. In addition to querying the database, we can see the slow trace in that SERVICE1 sends a request to SERVICE2. That’s useful info for sure. Let’s take a look more closely at the slow trace.

In the __Trace Graph view__, every node represents a Span, and on the left-hand side, we can see the percentage of time with respect to the total trace duration that the subgraph that has this node as root takes. So by inspecting this, we can see that the branch representing the HTTP GET from SERVICE1 to SERVICE2 takes most of the time of the span. So it seems the main suspect is SERVICE2. Let’s take a look at the Metrics now. They might give us more information. If we look at the thread.elu, we can see that for SERVICE2, it went 100% for some seconds. This would explain the observed behavior.

So now, going to the SERVICE2 code route, we can easily spot the issue. We were performing a __Fibonacci operation__. Of course, this was easy to spot as this is a demo, but in real scenarios, this would not be so simple, and we would need some other methods, such as CPU Profiling, but regardless, the info we collected would help us narrow down the issue quite significantly.

So, that’s it for the demo. We’ve created a repo where you can access the full code, so go play with it! 😎

Main Takeaways

Finally, we just want to share the main takeaways about implementing observability with Open Software Tools:

Setting up observability in our Node.js apps is actually not that hard.
It allows us to observe requests as they propagate through a distributed system, giving us a clear picture of what might be happening.
It helps identify points of failure and causes of poor performance. (for some cases, some other tools might also be needed: CPU profiling, heap snapshots).
Adding observability to our code, especially tracing, comes with a cost. So Be cautious! ☠️But we are not going to go deeper into this, as it could be a topic for another article.

Before you go

If you’re looking to implement observability in your project professionally, you might want to check out N|Solid, and our ’10 key functionalities’. We invited you to follow us on Twitter and keep the conversation!