Virtualization Administration

Every server you have administered so far was handed to you. Somebody else decided how much memory it had, which network it sat on, what its disk was really made of, and whether anything was backing it up. This course puts you on the other side of that handover: you operate the host, and the machines are yours to create, size, isolate, protect and hand over. You work on **virt01**, your own Ubuntu Server 24.04 LTS machine, running QEMU/KVM with libvirt — the foundation underneath Proxmox VE, and the stack Ultiblob's own private cloud is built on. Proxmox VE, VMware and Hyper-V are in the course as the professional context: what each one adds, what it names things, and when each is chosen. What you build with your own hands is the layer they all sit on. Over ten lessons you: - work out what your host can actually do, and tell hardware acceleration from software emulation honestly; - read the management plane — daemons, configuration, logs, task history — and make it log what you need; - define virtual machines from XML and from `virt-install`, boot them, watch their consoles, and build a cloud-init seed the way a cloud image expects one; - run OS containers with systemd-nspawn and machinectl, with real CPU and memory limits, and choose between a container and a virtual machine on evidence; - lay out storage pools and volumes, understand qcow2 against raw and thin against thick, make a template with linked clones, and use snapshots as the rollback for a change you have written down; - build isolated guest networks with DHCP and reservations, prove the isolation, then connect what the design says should be connected; - back a guest up, apply a retention policy, destroy it and restore it in a timed drill you record; - delegate one guest to an operator account with libvirt access control, and prove it cannot touch anything else; - design a three-node cluster with quorum, fencing and HA for a customer, and run the maintenance drill a single host can actually do; - operate the host: updates, reboot planning, metrics, logs, capacity arithmetic, and two seeded faults you have to diagnose and fix from the symptom. The final project is an engagement for Ferncourt Veterinary Clinic, an invented customer: you prepare one host for their applications — storage, isolated networks, a virtual machine and a container, a proven backup, a least-privilege operator account, a runbook and a design note — and fix three faults seeded into it before you hand it over. This is a learning pathway toward systems administrator and platform engineer roles. It awards an Ultiblob Certificate of Completion; it is not a vendor certification and makes no promise about employment.