observabilitycon logo
Event hero background image

How Just Eat Takeaway.com load tests with Grafana Cloud k6 to ensure millions of people get their orders reliably

How Just Eat Takeaway.com load tests with Grafana Cloud k6 to ensure millions of people get their orders reliably

Tip: Click the gear icon in the bottom right of the player to adjust your preferred viewing settings. If you have any issues, reach out to us at events@grafana.com.

Slack
Connect with the community on Slack

Company: Just Eat Takeaway.com

Industry: Software & Technology

Just Eat Takeaway.com (JET) is a leading Amsterdam-based online food delivery company, operating popular brands such as Just Eat, Takeaway.com, Lieferando, and SkipTheDishes across more than 15 countries. As JET expanded through rapid growth and major acquisitions, its engineering teams faced increasing complexity unifying systems across diverse markets, underscoring the need for a centralized, reliable observability and monitoring strategy.

Challenge

Just Eat Takeaway.com (JET) processes over half a billion orders per year across 17 countries and 15 time zones, serving 60 million active customers and 350,000 partners. With such massive global demand, even small performance degradations can ripple through and impact the ability for users to order food.

However, a major production incident exposed a blind spot in their testing strategy. During a Friday night peak period, a “non-critical” Tier 3 analytics service migrated to EKS without proper load testing. When one EKS node hit 100% CPU, shared gateways throttled Tier 1 and Tier 2 services, causing a significant outage and drop in customer orders.

Solution

Following the incident, JET took these learnings and re-architected its load testing strategy and adopted Grafana Cloud k6 as its global load testing platform, replacing their existing in-house “Rambo” tool. To migrate fully over to Grafana Cloud k6, JET took: 

  • A user-centric approach: Engineers across teams tested multiple tools before choosing k6 themselves, leading to high adoption and excitement.
  • Integrated in CI/CD: Added a “Run k6” step in the build pipeline, allowing teams to automatically validate performance before deployment to ensure customers get the best experience possible. 
  • Scheduled & on-demand testing: Weekly analytics stress tests simulate 45 million requests in an hour (≈17,500 requests/sec). And order flow tests simulate the full customer journey (home → menu → basket → checkout) across all regions.
  • Testing in production: Teams run all tests directly in production to build real-world confidence.

“We are not load testing to celebrate every success. We're not testing to celebrate every green light we get from a test. No, we load tests to uncover failures. And in this incident I just shared, it would've been prevented if it had been load tested before reaching 100% rollout.”

— Christopher Orchard, Senior Site Reliability Engineer

Impact

Grafana Cloud k6 has transformed load testing from a maintenance task into a core pillar of JET’s reliability engineering, ensuring that millions of customers can order their food, groceries, and every item they need, confidently and without disruption, even on the busiest Friday nights.

  • Faster detection of regressions: Tests in CI/CD catch performance issues before production.
  • Confidence in production reliability: Testing under real-world conditions ensures peak-time readiness and teams feel empowered to run schedule tests to get continuous feedback.
  • Curated dashboards: Grafana Cloud k6 offers curated dashboards with extensive test run information with the ability to dig deeper into how an individual report or endpoint is performed. 
  • Prevented repeat incidents: Continuous load testing ensures critical flows like order placement remain resilient.
  • Scale validation: Able to confidently handle tens of thousands of requests per second across global regions.

“All of this we achieve using Grafana Cloud k6. And out of the box, we get a lot of very useful features. We get given curated dashboards with extensive test run information.”

— Christopher Orchard, Senior Site Reliability Engineer

Speakers