Greetings.
I’ve recently started experimenting with MAAS and Juju together as a method for automated deployment of k8s clusters in my home lab.
I have this working however the time taken to fully deploy the charmed-kubernetes bundle feels a bit excessive to me, taking around 45 minutes in total (which includes an initial 10 mins of physical machine provisioning by MAAS). Watching Juju status during that time, it appears that installation and config of the kube-api server application alone takes around 10 minutes and calico isn’t much faster.
Teardown (juju destroy model) is fairly underwhelming in its performance too, 10 mins and upwards and usually getting stuck on removing the kubeapi-load-balancer application and thus requiring a subsequent --force.
My question here is, does this seem ‘normal’ / expected to anyone else with a similar setup?
As a frame of reference;
-
machines here are all bare metal. Dell PowerEdge M630 with 12 physical cores and 64GB RAM
-
MAAS and Juju controllers are on separate, physical servers of the same type and spec
-
10GbE networking
-
MAAS v3.4.2 on Ubuntu 24.04
-
Juju v3.5.1 on Ubuntu 24.04
-
no modifications or overlays, just a simple “juju deploy charmed-kubernetes” as per the quick-start tutorial
Thanks in advance for any thoughts and advice. The solution seems a good one, save for having to wait 45 minutes to deliver a working cluster.