There were new releases this weekend to fix breaking Hyperkube (more below). Note that Hyperkube is deprecated and will not be built by the Kubernetes project for 1.19, so if you’re the community images for your builds, you need to come up with another plan.
Per last week’s LWKD update, the image repository move was rolled back and will need to be retried later.
Next Deadline: Docs complete July 6, Code Freeze July 9
Hopefully you’re finishing your docs now, so USians can take the holiday weekend off.
1.18.5, 1.17.8, and 1.16.12 were released June 26 to fix multiple kube-proxy failures in Hyperkube. If you don’t use Hyperkube and already installed the prior bugfix release, you don’t need to care. If you do use Hyperkube, skip last week’s patch release and go straight to this one (but be prepared to take on your own Hyperkube maintenance pretty soon)
- Replace the built-in pod prememption logic with a defaultpreemption plugin
- Smoosh reserve and unreserve scheduler plugins together, and unreserve things if you can’t reserve them all
- Metrics: kubelet_running_containers and kubelet_running_pods “lose _count”, and multiple scheduling metrics get a profile labels
- Stop trying to use the docker config cache if it doesn’t exist
- It’s easier to upgrade custom component configs in kubeadm
--kernel-memcg-notificationmoves to the kubelet config file
- Fix a couple of scheduler API encoding mistakes to avoid case-sensitivity issues
- Make resource versions in lists consistent regardless of whether API requests are paginated or not
- Do some DNS1123Label validation
Original Source: http://lwkd.info/2020/20200630