Since its inception in 2004, Lansweeper has been at the forefront of helping businesses understand, manage, and protect their IT devices and networks through a powerful IT asset management platform. As the platform grew from an on-premises solution to a cloud-based SaaS offering, Lansweeper expanded its reach to a global, multi-region customer base. With this growth came the inevitable challenges of scaling observability and ensuring the engineering team could maintain performance and reliability across regions.

Observability means you know what’s happening in your software systems, because they tell you. They tell you with telemetry: data emitted just for the people developing and operating the software. You already have telemetry–every log is a data point about something that happened. Structured logs or trace spans are even better, containing many pieces of data correlated in the same record. But you want to start from what you have, then improve it as you improve the software.

As discussed in the first article in this series, a Center of Production Excellence (CoPE) is a more or less formal, provisional subsystem within an organization. Its purpose is to act from within to change that organization so that it’s more capable of achieving production excellence. The series has, to date, focused mainly on how best to construct such a subsystem and what activities it should pursue. In this concluding post, however, I want to return to the point of a CoPE, discuss signs of success, and evaluate the impacts it’s having.

Let's be real, we've never been huge fans of conventional unstructured logs at Honeycomb. From the very start, we've emitted from our own codestructured wide events and distributed traces with well-formed schemas. Fortunately (because it avoids reinventing the wheel) and unfortunately (because it doesn't adhere to our standards for observability) for us, not all the software we run is written by us. And Kubernetes is a prime example of such a load-bearing part of our infrastructure.

At Honeycomb, we know how important it is for organizations to have a unified observability platform. This is why we’re launching Honeycomb Telemetry Pipeline and Honeycomb for Log Analytics: to enable engineering teams to send and analyze data—including logs—into a single, unified platform. 

Over the past six weeks, we introduced a series of impactful updates aimed at making your observability workflows faster, more unified, and more collaborative. Here’s a snapshot of what we worked on.

Frontend development has evolved rapidly over the past decade, but one challenge remains constant: understanding what’s happening in real-time across diverse browsers, environments, and user interactions. This is where observability steps in—but how does it apply to the frontend world where user experience can break in countless, unexpected ways?

Real user monitoring (RUM) began as a straightforward approach to tracking basic web performance metrics. Focused on things like page load times and response rates, RUM relied on server-side logging and simple browser timings. While these tools captured Core Web Vitals (CWVs), they offered limited insights into how users actually interacted with pages, focused mainly on server-side performance.

Oh no, I’m getting out-of-memory errors! How much memory is my app using? To find out, we go look for a metric that tells us how much memory is available, and we graph it around the time that our errors occurred.

Recently, we announced the launch of Honeycomb for Frontend Observability, our new solution that helps frontend developers move from traditional monitoring to observability. What this means in practice is that frontend developers are no longer limited to a metrics view of their app that can only be disaggregated in a few dimensions. Now, they can enjoy the full power of observability, where their app collects a broad set of data as traces to enable much richer analysis of the state of a web service.

Refinery is Honeycomb's sampling proxy, which our largest customers use to improve the value they get from their telemetry. It has a variety of interesting samplers to choose from. One category of these is called dynamic sampling. It's basically a technique for adjusting sample rates to account for the volume of incoming data—but doing so in a way that rare events get more priority than common events.

We’ve posted before about how engineers on call at Honeycomb aren’t expected to do project work, and that whenever they’re not dealing with interruptions, they’re free to work on whatever will make the on-call experience better.