The Canonical Distribution of Kubernetes: Development Summary #4

This blog was originally posted by Tim Van Steenburgh

July 21st concluded our most recent development sprint on the Canonical Distribution of Kubernetes (CDK). Here’s a look at what we did.

Fixes and Improvements

Check out the full list on GitHub. Here are some notables:

• Made load balancer port configurable
• Changed default --service-cluster-ip-range to a /16 CIDR to allow more NodePort IP addresses
• Fixed etcd snapshot action
• Increased default worker node constraints to 4 cpu, 4GB RAM

Testing

• Added a test to ensure dashboard is operational after deploy
• Added a test for the built-in microbot example
• Added a Jenkins job to test master charms with stable snaps. When this is green it means we can release whatever new fixes/features we have queued up in the charms, giving us the confidence to do more frequent releases.

Features

• Calico spike. We want to provide a CDK + Calico deployment option that works on any cloud, just like our CDK + Flannel option. We’ve decided to go with a Calico-on-Flannel (Canal) approach initially. Canal combines the network policy enforcement of Calico with the ease-of-deployment of Flannel. Work begins in the current sprint!
• RBAC spike. We mapped out the work necessary for enabling RBAC via charm config. Work begins in the current sprint!
• Updated the canonical-kubernetes-elastic bundle. This bundle has been added to our Jenkins build process and updated with the latest 1.7 charms.