Connecting Cloudflare and Honeybadger with OpenTelemetry

Honeybadger Insights is pretty great. It's even greater when you connect it to Cloudflare and start shipping logs directly from Cloudflare Workers to Insights. If you're already using Cloudflare to distribute your applications, it's easy to set up.

Cloudflare Workers logs capture what your application is doing at the edge: console.log() output, uncaught exceptions, and system-generated logs for each request. That data can be very important for your application's health — especially for web apps in distributed systems. Insights gives you the tools to act on it.

In this article, we'll use OpenTelemetry to connect Cloudflare and Honeybadger. For that, we need a test app. We'll make one using RedwoodSDK.

RedwoodSDK is a server-side React framework described as "a framework for humans." It's a breeze to spin up if you already have Node.js installed — with just a few commands, you can get a server up and running.

Deploying it is just as easy. Once you have a Cloudflare account ready, push everything to production with npm run release.

As specified in RedwoodSDK's documentation, I had to complete the additional step of creating a workers.dev subdomain, but it still worked on the first try. No bugs or troubleshooting needed.

Why you should connect Cloudflare to Honeybadger Insights

Cloudflare already has an observability page, so for small web applications, you could stop there. For bigger applications, Cloudflare's observability page may not be enough, since its tooling is relatively shallow. It splits your data between logs and traces, making it harder to find and fix errors. Additionally, the query builder is fragmented, so it's hard to save visualizations for repeat access.

Honeybadger Insights makes log analysis that much easier. Insights has everything you need: a powerful query system, natural language search, easy visualization generation, and configurable alarms. Forget separating data into logs and traces — Insights uses wide events, each one a collection of properties bundled into a single record. User behavior, business intelligence, system logs, and Cloudflare Workers logs can all be sent to Insights.

OpenTelemetry: the glue that makes it all work

OpenTelemetry (OTel) is an open-source observability framework for cloud software. Insights supports the OpenTelemetry Protocol (OTLP), including traces, metrics, and logs; for this example, we'll focus on logs. Like most OTLP log destinations, Insights uses an endpoint ending in /v1/logs: https://api.honeybadger.io/v1/logs. We'll use this for our guide, but you'll also need an API key, which can be generated in your project settings.

Cloudflare also supports OTel, specifically traces and logs. We're interested in the logs, which Cloudflare describes as "...application logs, including console.log() output and system-generated logs." You can read more about their logging methods in Cloudflare's documentation.

Now let's hook it all up.

Setting up log exports

The process starts once you've created a RedwoodSDK app and deployed it to Cloudflare. In the sidebar of your Cloudflare account, find the Observability tab under the Compute dropdown.

Scroll down from Account Home to find the Compute section.

The page will show an events timeline that can be configured for live log tailing, searching, or sorting.

Navigate to the Destinations sub-page. Now, it's time to create a destination.

Click Add destination, then switch over to the Logs tab. Next, head back to your Honeybadger account. Each destination sends logs to a single Honeybadger project, and a project's API key is located under the Settings tab in the project's top menu. From there, navigate to API keys and generate a new key. You can label it, too — I named mine cloudflare.

Now armed with an API key, we can return to the Cloudflare page and start filling it out. In the input field for the OTLP Logs Endpoint, copy and paste https://api.honeybadger.io/v1/logs as the address. Name the destination whatever you want, then add a custom header.

For the custom header name, enter X-API-Key, and for the value, paste your API key. The header name must be exactly X-API-Key, or Honeybadger will reject the logs. Once you click Save, Cloudflare will run an internal check. If everything is filled in correctly, the destination will save successfully.

Voilà! We have a working log pipeline between Cloudflare and Honeybadger. Now we can monitor our Worker's activity much more easily.

Integration testing

Cloudflare may say that our log pipeline works, but we need to be sure we're receiving logs. Let's define a log line, trigger it, and then make sure we're ingesting it.

I added this route to my src/worker.tsx file, in the route section:

route("/hello", function () {
  console.log("Hello world!");
  return new Response("logged");
}),

If you've followed RedwoodSDK's quick start, the route section should look like my example below.

render(Document, [
  route("/", () => new Response("Hello, World!")),
  route("/ping", function () {
    return <h1>Pong!</h1>;
  }),
  route("/hello", function () {
    console.log("Hello world!");
    return new Response("logged");
  }),
]),

We can now prompt our Worker to generate a "Hello world!" log. I ran curl -s https://<your-worker-url>/hello and then checked Cloudflare to make sure the message was recorded before moving over to Insights.

Events won't appear in Insights right away — give Cloudflare a few minutes to ship them over. Check it out!

Clicking on the Insights tab brings up the default view, and since this is a dummy project, I have no other results. If you had other integrations sending data, you could filter by event_type, or filter on a specific message like "Hello world!" For logs, severity is an excellent filter, too.

Everything else Insights can do

Now with Cloudflare and Honeybadger connected, our logs will flow into Insights, where they can be sorted and measured. We can even turn them into widgets to track errors, API activity, and more. The more you integrate with Insights, the more you can observe and act on. Connecting your Workers to a monitoring service is always a good idea, allowing you to catch problems before they affect your customers.

Why not use Honeybadger for monitoring? It's easy to install and set up, no matter what framework you're using. Try it — it's free to make an account.