What’s new in MicroK8s v1.23?
Committed to our mission we released today MicroK8s v1.23, only a few hours after the upstream Kubernetes came out. For those who may not already know MicroK8s is a low-ops, minimal production Kubernetes distribution suitable for:
- Developer workstations
- IoT
- Edge
- CI/CD
Give the 1.23 release a spin with:
sudo snap install microk8s --classic --channel=1.23You can read more on what the latest Kubernetes offers in the official announcement. Here we focus on the work and main features we put on this MicroK8s release.
Worker only nodes
Up until now all nodes in a cluster were running the control plane contributing in this way to the the resilience and fault tolerance of MicroK8s. You now have the option to add worker only nodes to a cluster. This kind of nodes do not run the K8s control plane and do not contribute to the high availability of the cluster. They however consume less resources and thus are suitable for low end devices. Worker only nodes are also suitable in configurations where the nodes running the K8s workloads are unreliable or they are just not trusted with the task of holding the control plane. You can read more on forming clusters with or without worker nodes in our docs.
GPU operator with Multi-instance support
The updated NVIDIA GPU operator we ship with the gpu add-on, microk8s enable gpu , enables us to configure Multi-instance GPUs via config maps. Top of the line, beading edge technology by the NVIDIA engineers!
FPGA, first steps
Really excited to test the waters in the FPGA lake with the new InAccel add-on. Join us with: microk8s enable inaccel
Hassle free lxc deployments
In this release MicroK8s will detect if it is deployed inside lxc containers and will instruct kube-proxy not to configure connection tracking. The connection tracking setup is inherited from the host node.
Lots of add-on upgrades
We would like to thank our community members and especially @balchua for actively maintaining the add-ons in the distribution.
Dqlite improvements
That is something you may never notice, yet the good work on our default datastore, dqlite, progresses nicely. In addition to the performance and reliability improvements of the core dqlite, we also split dqlite in its own systemd process to improve stability.
Charmed MicroK8s
MicroK8s officially enters the world of Juju charms. We have taken the MicroK8s charm under our wings. Head over to charmhub.io to give it a try.
Summing up
The 1.23 release is the best release so far. Lots of details have been ironed out. I am sure you will enjoy it. A big shout out to all of you who contributed to this release. @balchua, @balasu, @gkarthiks, @jlettman, @erulabs, @MichaelCduBois, @farazmd, @BabisK, @sfstpala, @tobiasmuehl, @niladrih, @jlettman, @eliaskoromilas, thank you!
Looking forward to meeting you at the Kubernetes Slack, in the #microk8s channel.
