carwow connects auto dealers and buyers in a transparent marketplace, creating better buyer experiences and incremental sales for dealers.
- Microservices-lite, one app per region, multiple apps per country
As carwow’s successful online car review and selling business expanded into larger markets in new regions, their engineering team needed deep, system-wide visibility into the performance impacts of interactions among the components of their rapidly evolving infrastructure.
In scaling their systems to support growing traffic from one app to many apps and background jobs and events, carwow’s engineering team has gained much-needed flexibility and agility, but as all systems architects and engineering leaders understand, increasing complexity leads to greater opacity:
- Because the organization is growing quickly, it’s becoming more difficult for one team of developers to have a good understanding of how the code they ship will interact with code shipped by other teams
- Because the system is more loosely-coupled (which provides flexibility) than before, it’s easier for new code to trigger jobs that have an escalating, ripple effect through the system
What They Needed
- An observability service that allowed them to maintain deep visibility into their infrastructure as they scale it to meet their flexibility requirements
- A single interface to trace request performance and the interaction of systems across their architecture from beginning to end
Honeycomb @ carwow
Among the components in their infrastructure is Kafka, a software stream-processing platform. In a typical use case at carwow a user visits their website and takes an action, which publishes one or more jobs into the Kafka system. Other apps in the carwow environment then pick up the jobs and may themselves make multiple API requests to other applications, which may in turn do the same–and the process can continue onward into the system, sometimes looping back to the same applications as needed to fulfill the original request.
When there’s an issue, New Relic can show the jobs running, the API requests but can’t tell us the source of this thing that’s happening right now, can’t trace it back to a request, an earlier job in the chain.
With access to the high-cardinality querying of wide events and deep tracing introspection supplied by Honeycomb, carwow can now find a request from a user, what messages it sent to Kafka, what jobs were queued, and rapidly arrive at the one request that caused a particular job to fire.
Before Honeycomb, we couldn’t really observe what was going on across the entire system. Honeycomb Tracing makes everything transparent again.
The engineers at carwow now know that when they see a problem in one system and see that it crosses multiple systems, they can use Honeycomb to follow the request through its entire life-cycle and understand the impact on different subsystems in the code. Honeycomb enables and turbo-charges cross-team collaboration to solve issues.
We now regularly solve issues being discussed in Slack just by pasting a link to a Honeycomb trace diagram.
Having this visibility and collaborative power has allowed the carwow engineering organization to work together more effectively; they are able to quickly and easily find and eradicate inefficiencies and performance problems introduced by code making requests across components developed by different teams.
Honeycomb has really opened our eyes to how our system behaves, how we can make it more efficient.
As carwow continues to scale their rapidly growing, successful enterprise, they look to Honeycomb to help them maintain the stability they require to deliver a quality service to their users alongside the flexibility they need to deliver it.
Tracing is a game-changing thing for us. We loved Honeycomb before, but this has sped up our ability to observe how our entire system behaves.
April 2, 2018
Honeycomb at Tapjoy: Faster Time to Confidence with Observability
Tapjoy's engineers get greater confidence and save days of effort when cutting over services for their attribution partners using Honeycomb for visibility into their monolithic Rails setup.
April 1, 2019
Always Bee Tracing
At Honeycomb, we practice observability to know how our production systems are behaving in real-time. Distributed tracing is just one feature we use to solve issues fast and see what’s happening across all services. With tracing inside Honeycomb, we switch easily to different chart views – histograms, heat-maps and waterfall trace views.
August 3, 2021
Distributed Tracing – A Guide for Microservices and More
Want to know what tracing is, what it's for, and how to get the power that it offers to debug and explore your code in production? Distributed Tracing – A Guide for Microservices and More is for you. The 3rd volume in the Honeycomb Observability e-Guide Series, this guide covers an introduction to the topic of tracing, its history, how to think about tracing, and how to add traces to your instrumentation.