Windows Worker Reorganisation

[mtelvers]

Two physical hosts are running the Windows workers and other workloads: Odawa and Hopi.

Odawa is running Proxmox with individual VMs created via the command line. Windows is installed via Microsoft Deployment Toolkit (MDT) and then customised by an Ansible playbook, which performs the build of the ocluster/obuilder code.

Hopi is running KVM/LIBVIRT. We have packer templates which build Vagant Windows boxes. These can then be quickly deployed with vagrant up and subsequently configured by Ansible (as for odawa). However, Hopi only has Windows 2022 VMs and no Windows 2019 VMs. These can't be added as there is insufficient disk and memory capacity.

Over time, the deployments have diverged; Odawa has been virtually frozen (other than regular Windows updates), whereas Hopi is running the latest version of ocluster worker.

• [x] Create enough free capacity on Hopi to allow adding a single Windows 2019 VMs
• [x] Migrate management scripts from Odawa to Hopi
• [x] Reinstall Odawa with Ubuntu 22.04 LTS
• [x] Configure Odawa to run KVM, Vagrant, Packer etc.
• [x] Install Windows 2022 and Windows 2019 workers on Odawa
• [x] Remove workers from Hopi leaving exclusively for other workloads
• [x] Setup Grafana to monitor the VMs on both hosts

[mtelvers]

This work has now been completed.