Use when operating KVM/libvirt virtual machines or LXD system containers, including lifecycle, snapshots, backups, storage, networking, and host inspection. Use linux-container-engine and linux-container-deployment for Docker or Podman application containers.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Use when operating KVM/libvirt virtual machines or LXD system containers, including lifecycle, snapshots, backups, storage, networking, and host inspection. Use linux-container-engine and linux-container-deployment for Docker or Podman application containers.
Two-family skill. KVM/libvirt (virsh, virt-install) is portable across both
families; LXD system containers are Ubuntu-centric (Canonical/snap) with no
native LXD on RHEL. Body uses Debian/Ubuntu; substitute per this matrix.
Application containers (Docker/Podman) moved out of this skill. For the
container engine, running containers, and image cleanup, use the
12-containers-and-orchestration skills: linux-container-engine
(install/configure Docker & Podman), linux-container-deployment
(run containers, compose, systemd/Quadlet), and linux-image-hygiene
(prune images/volumes). This skill keeps KVM/libvirt VMs and LXD system
containers only.
Concept
Debian/Ubuntu
RHEL family
System containers
LXD (snap)
systemd-nspawn (no native LXD); Podman for app containers
Managing LXD system containers or KVM/libvirt VMs on a host.
Investigating VM or LXD-container lifecycle failures.
Taking snapshots or backups before risky changes.
Do not use when
The task is application containers (Docker/Podman, compose, image cleanup); use the 12-containers-and-orchestration skills (linux-container-engine, linux-container-deployment, linux-image-hygiene).
The task is configuration management for the host itself; use linux-config-management.
The task is ordinary application service management outside the VM/LXD layer.
Required inputs
Artefact
Source
Required?
If absent
Host family, hypervisor/container layer, target identity, and desired lifecycle outcome
Operator and host inventory
required
Discover read-only; do not act on an ambiguous name.
Current guest definition, storage/network attachments, state, and host capacity
libvirt/LXD host
required for mutation
Return a preflight report only.
Backup/snapshot policy, downtime window, recovery target, and change authority
Workload owner
required for destructive or disruptive action
Stop before mutation.
Workflow
Identify the owning virtualization layer and target workload.
Inspect current state before changing it.
Apply the matching workflow below for lifecycle, snapshot, backup, or startup diagnosis.
Verify the guest or container state and host-level impact after the action.
Stop if the target, storage/network ownership, capacity, backup consistency, rollback, downtime, or action authority is unresolved.
Recover by reverting the recorded definition/snapshot or restoring the independent backup, then verify both guest state and workload health.
Quality standards
Keep host and guest responsibilities distinct.
Snapshot or back up before risky mutations when possible.
Validate both orchestration state and actual workload health.
Anti-patterns
Treating LXD system containers and Docker/Podman application containers as interchangeable. Fix: route application containers to category 12.
Deleting a guest, disk, snapshot, or image before proving ownership. Fix: inventory attachments, dependants, backup, and retention first.
Debugging only inside the guest. Fix: inspect host capacity, libvirt/LXD state, storage, networking, and logs.
Calling a snapshot a backup. Fix: export or copy recovery data to an independent failure domain and test restore metadata.
Starting a migration or resize without capacity/downtime checks. Fix: preflight destination compatibility and define rollback.
Outputs
Artefact
Consumer
Acceptance condition
Guest diagnosis or lifecycle action
Virtualisation operator
Correct target/layer is identified and host plus guest state match the requested outcome.
Snapshot/backup/recovery record
Workload owner
Artifact location, consistency method, retention, restore command, and verification are recorded.
Capacity and impact evidence
Infrastructure owner
CPU, memory, storage, network attachments, downtime, and rollback are checked before change.
For Docker/Podman application containers: linux-container-engine, linux-container-deployment, and linux-image-hygiene in 12-containers-and-orchestration.
Evidence Produced
Artefact
Acceptance condition
Virtualisation evidence
Includes target definition, host capacity, attachments, state/logs, backup or snapshot identity/location, lifecycle result, and workload health.
Capability contract
Inspection defaults to read-only. Read/execute access to the host is required. Start/stop, snapshot, backup, resize, migration, XML/profile edits, storage/network changes, or deletion require explicit authority. Deletion and overwrite require named target verification and a recoverable backup.
Degraded mode
Without host or guest access, review supplied definitions/logs and mark capacity, storage consistency, runtime, and workload health not assessed. Without a verified backup or recovery path, stop before destructive or irreversible work.
Decision rules
Choice
Action
Failure or risk avoided
Full OS isolation/legacy kernel workload
Use KVM/libvirt
Container-host coupling.
Lightweight system container on supported Ubuntu host
Use LXD
Unneeded VM overhead.
Packaged application container
Route to category 12
Competing lifecycle tools.
Worked example
Before changing a production libvirt VM, identify its disks and networks, check host capacity, create an application-consistent backup to independent storage, record restore steps, take a short-lived snapshot, apply the approved change, verify guest and application health, then remove the snapshot only after the retention gate.
This skill is self-contained. Every command below uses standard tools
(lxc, virsh); the body shows Debian/Ubuntu, with RHEL-family substitutions
(SELinux :z/:Z) per the Distro support matrix above. The sk-* scripts
in the Optional fast path section are convenience wrappers — never required.
This skill owns the VM and system-container layer on a host: LXD system
containers (Canonical's native, Ubuntu-centric) and KVM/libvirt for full
virtual machines.
It does not own:
Application containers (Docker/Podman), compose, and image cleanup — use
the 12-containers-and-orchestration skills: linux-container-engine,
linux-container-deployment, linux-image-hygiene.
Kubernetes — out of scope for v1.
Cloud provider VMs (EC2, DigitalOcean droplets) — managed at the
provider side.
Applications inside containers — managed by application-specific
skills.
Informed by the Canonical Ubuntu Server Guide (LXD, KVM chapters).
When to use
Listing LXD containers and VMs on a host.
Creating or restoring LXD container snapshots.
Exporting/backing up an LXD container or VM to a tar file.
Debugging why an LXD container or VM won't start.
When NOT to use
Running or deploying Docker/Podman application containers — use
linux-container-deployment (engine setup: linux-container-engine).
Deploying a web application inside a container — use
linux-site-deployment.
Host firewall rules — use linux-firewall-ssl.
Kubernetes/Nomad orchestration — out of scope.
Standing rules
LXD over legacy LXC.lxc (the LXD CLI) is the modern, declarative
wrapper. Direct lxc-* (legacy) commands are discouraged.
Always snapshot before mutating. Snapshot → mutate → verify. Roll
back on failure.
Snapshots are not backups. They live on the same pool. Export
containers/VMs to tar on separate storage for real backup.
Resource limits are mandatory in production. Unbounded containers
eat the host. Set limits.memory, limits.cpu, and limits.disk in
every profile.
Privileged containers are banned unless justified in writing.
Unprivileged is the default.
Quick reference — manual commands
LXD
# List all containers
lxc list
lxc list -f compact # compact format
lxc info <name> # detailed state, IP, resource use# Launch and basic lifecycle
lxc launch ubuntu:24.04 web01
lxc exec web01 -- bash
lxc stop web01
lxc start web01
lxc delete web01 --force
# Snapshots
lxc snapshot web01 before-upgrade-2026-04-10
lxc info web01 | grep -A10 Snapshots
lxc restore web01 before-upgrade-2026-04-10
lxc delete web01/before-upgrade-2026-04-10 # delete a snapshot# Full export / backup
lxc export web01 /backups/lxd/web01-$(date +%Y%m%d).tar.gz
lxc import /backups/lxd/web01-20260410.tar.gz
# Resource limits (applied to running container)
lxc config set web01 limits.memory 1GB
lxc config set web01 limits.cpu 2
lxc config device override web01 root size=10GB
# Copy between hosts
lxc remote add remotehost <ip>
lxc copy web01 remotehost:web01-copy
Full LXD reference (lxd init, profiles, storage backends, networks,
cloud-init integration, 6 worked examples) — see
references/lxd-reference.md.
Docker / Podman application containers are no longer covered here. Use the
12-containers-and-orchestration skills: linux-container-engine,
linux-container-deployment, and linux-image-hygiene.
echo"=== LXD ==="
lxc list 2>/dev/null || echo"(lxd not installed)"echoecho"=== KVM ==="sudo virsh list --all 2>/dev/null || echo"(libvirt not installed)"
(For Docker/Podman containers on the host, use sk-container-ps from
linux-container-deployment.)
Workflow: "Snapshot before I mess with it"
lxc snapshot web-prod before-upgrade-$(date +%Y%m%d-%H%M)
# ... do the risky thing ...# If it breaks:
lxc restore web-prod before-upgrade-20260410-1530
# When confirmed stable, delete the snapshot:
lxc delete web-prod/before-upgrade-20260410-1530
lxc info web01 | grep -A10 -E 'Status|Log'
lxc info --show-log web01
# Resource limits too tight?
lxc config show web01 | grep -E 'limits\.'
(For "why won't this Docker/Podman container start?" use
linux-container-deployment.)
Troubleshooting / gotchas
LXD containers "lose" their IP on boot. They use DHCP from lxdbr0
by default; if the bridge has no DHCP server, set a static IP in the
profile or container config.
KVM VMs with bridged networking don't get IPs. Check that the host
bridge has a DHCP server on the VLAN, or configure the VM with cloud-init
static networking.
LXD container file ownership looks weird from the host. Unprivileged
containers remap UIDs (uid 1000 inside = uid 1001000 outside). Use
lxc file push/pull to move files without fighting ownership.
LXD containers with state, IP, memory, disk, uptime
sudo sk-lxd-list
Create / list / restore LXD snapshots
sudo sk-lxd-snapshot
Full LXD container export to tar
sudo sk-lxd-backup --container <n>
These are optional wrappers. The lxc and virsh commands above are the
source of truth. For Docker/Podman, see sk-container-ps /
sk-container-prune in 12-containers-and-orchestration.
Scripts
This skill installs the following scripts to /usr/local/bin/. To install:
sudo install-skills-bin linux-virtualization
Script
Source
Core?
Purpose
sk-lxd-list
scripts/sk-lxd-list.sh
no
All LXD containers with state, IPv4/IPv6, memory and disk usage, uptime.
sk-lxd-snapshot
scripts/sk-lxd-snapshot.sh
no
Create, list, or restore LXD container snapshots with a naming convention.
sk-lxd-backup
scripts/sk-lxd-backup.sh
no
Full LXD container export to tar, with restore metadata, for off-host backup.
