The New York Times’ is a US media company serves digital journalism to millions of visitors every day. The format of our stories is constantly experimented with; for example we publish graphics based on election data ingested from APIs, question and answer led discussions, breaking news live coverage, and quizzes. This leads to a lot of applications.
Our previous experience with infrastructure may be a familiar one: an unruly number of virtual machines, which led us to containers. Containers give our web developers who are not infrastructure engineers the opportunity to configure and launch their applications with little oversight.
Kubernetes offers us an infrastructure for our numerous applications at scale. Leveraging the Kubernetes API, we’ve built a self-service admin interface for developers (not sysadmins) to configure and launch their applications at scale, similar to the Kubernetes Dashboard project, tailored to our development workflow.
- Different story formats of the NYT
- Problems with our older VM infrastructure
- Moving to containers, benefits for both the sysadmin and non-sysadmin team
- Our brief usage of Fleet for application init and problems there
- Adopting Kubernetes and the benefits
- Network architecture of how we run apps in Kube, our Kube Router
- Demo of our custom dashboard for managing apps
- Our move towards Kube Dashboard