Zendesk’s Cloud‑Native Transformation: Migrating to Kubernetes and Microservices

Zendesk, founded in 2007 as a monolithic Rails application backed by MySQL, grew to over 145,000 paid accounts and needed to modernize its infrastructure. The legacy on‑premise data center caused slow deployments, high operational overhead, and inconsistent environments between staging and production.

Challenge: Scaling the single‑tenant Rails app was becoming untenable; the central ops team required long lead times for hardware provisioning, and extensive Chef scripts created divergent configurations across environments.

Solution: In late 2014 the engineering team evaluated container technologies, adopted Docker, and spent six months developing best‑practice microservice patterns. They created the “ZDI” (Zendesk Docker Integration) tool to spin up containers quickly, but soon realized they needed a proper orchestrator. By summer 2015 they selected Kubernetes, migrated workloads to about 15 AWS clusters, and by early 2017 had production traffic on Kubernetes. New services are now built directly for Kubernetes, resulting in roughly 70% of Zendesk’s applications running on the platform.

“Kubernetes seemed to be built to solve our problems.” – Jon Moter, Senior Principal Engineer, Zendesk

The team also enhanced their open‑source deployment tool Samson to integrate with multiple Kubernetes clusters, read YAML manifests from GitHub, and adjust replica counts, CPU, and RAM via a UI before calling the Kubernetes API. This reduced resource‑change cycles from days to 1‑2 minutes.

Operational benefits include rapid self‑healing during AWS outages (issues resolved in minutes without manual intervention) and a unified orchestration platform that standardizes tooling, monitoring, and alerting across globally distributed teams.

To help other organizations adopt Kubernetes, Moter advises building internal expertise, starting with small applications, and gradually scaling up, while recognizing the learning curve and the need for a dedicated team familiar with the platform.

Business Impact: Resource‑change time dropped from days to minutes, incident recovery now takes minutes instead of hours, and 70% of applications run on Kubernetes, delivering greater flexibility, speed, and reliability for Zendesk’s services.