That first model will simply involve 2 networks, the underlay network 192.168.3.0/24 Virtual Router and OpenVSwitch, to play in an isolated IP address space, behind my ISP router.
[ ] release an OpenStack with Ansible Kolla, using bare-metal inventory, of bare metal machines with podman / containerd / skopeo, instead of docker
[ ] Do the same as first point, but this time it is with OpenStack Neutron
[ ] next step, I provision kubernetes inside OpenStack, using terraform, and :
an ansible playbook with roles.
Kubeone and I want to try the openstack nodegroup autoscaler : I hit the app, and new cluster nodes are PXEless booted, and registered as cluster nodes. I stop hitting the app, and cluster nodes get drained and destroyed. It's like the garbage collector, but for our VMs/ClusterNodes.
I'll also try one rancher distribution, tha tis not k3s
and I'll try a multinodes multi master k3s.io along with k3d.io
[ ] rermount the whole thing, but this time with pulumi and a pulumi state storage backend of S3 type : inside Minio.
[ ] Try Ceph and GlusterFS as sorage drivers for Kubernetes.
[ ] fork inlets-operator, to make it work with a private OpenStack Kolla, see #5 (and the internet can be shut down, we won't care
I will set up a model for networking, while operating from bare metal.
192.168.3.0/24
Virtual Router and OpenVSwitch, to play in an isolated IP address space, behind my ISP router.k3s.io
along withk3d.io