Building out a managed Kubernetes service is a bigger job than you think

You might be working your code in containers. You may even have taken the plunge and orchestrated all of it with YAML code by way of Kubernetes. However infrastructure as code turns into a complete new degree of sophisticated when organising a managed Kubernetes service. 

On this sponsored episode of the Stack Overflow podcast, Ben and Ryan speak with David Dymko and Walt Ribeiro of Vultr about what they went by way of to construct their managed Kubernetes service as a cloud providing. It was a journey that ended not simply with a managed K8s service, but in addition with a wealth of extra tooling, upgrades, and open sourcing. 

When constructing out a Kubernetes implementation, you may summary away a few of the complexity, particularly should you use a few of the extra fashionable instruments like Kubeadm or Kubespray. However when utilizing a managed service, you need to have the ability to focus in your workloads and solely your workloads, which implies taking away the management aircraft. The consumer doesn’t have to care in regards to the underlying infrastructure, however for these designing it, the lacking management aircraft opens a complete heap of bother. 

When you take away this abstraction, your cloud cluster is handled as a single strong compute. However then how do you do upgrades? How do you keep x509 certifications for HTTPS calls? How do you get metrics? With out the management aircraft, Vultr wanted to speak to their Kubernetes employee nodes by way of the API. And wouldn’t you realize it: the API isn’t all that well-documented. 

They took it again to reveal requirements, the MVP function set of their K8s cloud service. They’d want the Cloud Controller Manager (CCM) and the Container Storage Interface (CSI) as core parts to have Vultr be a first-class citizen on a Kubernetes cluster. They constructed a Go consumer to interface utilizing these parts and figured, hey, why not open-source this? That led to some different open-source tasks, like a Terraform integration and a command-line interface. 

This was the beginning of a two-year journey connecting all of the dots that this mission required. They wanted a managed load balancer that would work with out the management aircraft or any of the instruments that interfaced with it. They constructed it. They wanted a quality-of-life replace to their API to meet up with all the things that at present’s developer expects: trendy CRUD actions, REST finest practices, and pagination. All of the whereas, they saved listening to their clients to ensure they didn’t stray too removed from the unique product. 

To see the outcomes of their journey, take heed to the podcast and take a look at Vultr.com for all of their cloud choices, obtainable in 25 areas worldwide. 

TRANSCRIPT

Tags: cloud computing, kubernetes, partner content, stackoverflow, the stack overflow podcast

More Posts