Case Study

LaunchDarkly Guesses Less, Knows More With Next-Gen APM

5 minute read


LaunchDarkly offers fast and reliable feature management for the modern enterprise.


  • AWS, ALB
  • Services written in Go, one especially large service
  • Large collection of legacy metrics and alerts written against Graphite



LaunchDarkly’s engineering team works hard to ensure that the features they ship meet the requirements and goals of their customers. In the past, they had success using an in-house Graphite instance to collect the metrics they used to monitor customer health, but as their business grew, they found it could not keep up with the increasing volume and cardinality of the data stream. They would know something was wrong, but could not get the performance or detail they needed to find out what it was or who it was affecting.

Graphite still met their needs for monitoring some aspects of system health, but they needed something better able to dig into how their users actually experienced their service. They evaluated additional vendors but did not find the performance-to-pricing balance they were looking for, until Honeycomb.

With Honeycomb, it’s not a case of keeping huge volumes of data in storage so your costs rise and you still need to keep more. Cost is predictable; you know that you can rely on getting access to that one event or piece of raw data you need, so it’s more reliable.

Mike Atkins, Software Engineer at LaunchDarkly

What They Needed

  • An observability service that allowed them to see user behaviors and feature adoption down to the individual customer level
  • The ability to continue using the legacy metrics tooling their team was comfortable with while also achieving the observability needed to dig into new issues they were facing as they scaled

Honeycomb @ LaunchDarkly

Once Honeycomb was in the mix, LaunchDarkly’s engineers identified which services were causing Graphite the most pain, and adapted their data streams to send the wider, more context-rich events they needed to Honeycomb instead. As a result, their Graphite server became less oversubscribed and better able to serve their day-to-day desire for metric dashboards.

Now, with Honeycomb available, LaunchDarkly’s team is solving the problems they couldn’t before. For example, they were recently able to identify and resolve an issue related to a bug in the AWS ALB that was affecting a subset of customers that use the LaunchDarkly service in a very specific way. Once the issue was discovered, they were further able to identify which users would encounter the bug based on their usage pattern and contact that customer to pre-empt a bad experience.

Before Honeycomb, we would just speculate wildly about who was impacted by a given issue, or what changes would affect which customer.

Mike Atkins, Software Engineer at LaunchDarkly

At Honeycomb, our goal is to ensure you can meet your business requirements for quality of service and customer happiness. Working alongside the tools you already use to augment their capabilities is part of our strength. With Honeycomb on your side, your use case and requirements for high-cardinality queries capacity will scale up, and we’ll be there to keep you successful and growing the whole time.

Honeycomb in one word: comprehensive.

Arun Bhalla, Engineering Manager at LaunchDarkly