Linux Virtualization Cloning a KVM Guest Virtual Machine

Internal

• Linux Virtualization Guest Operations

Overview

The goal of a cloning operation is to create an entirely new guest, based on the configuration of an existing guest. Entirely new storage must be provisioned during the cloning operation, and the O/S image must be transferred on the new storage and updated as described below. Also, care should be taken to avoid conflict while accessing shared resources - memory, CPUs, network devices.

Procedure

Shut Down the Source Virtual Machine

Shut down the guest to be cloned:

 virsh shutdown source_vm

virsh shutdown ocp36.basic-template

Export the Configuration

Export the configuration of the guest to be cloned:

 virsh dumpxml source_vm > .../vm-definition-repository/source_vm.xml

virsh dumpxml ocp36.basic-template > /root/environments/ocp36/ocp36.basic-template.xml

Adjust the Configuration

Copy the XML definition under a new name, conventionally the name of the guest being built.

cp ocp36.basic-template.xml ocp36.generic-template.xml

Edit the XML as needed. These are some things you may want to change:

Name of the Guest

Update the name of the guest as:

<domain type='kvm'>
  <name>the_new_name</name>
  ...

For more details on constraints associated with the guest name, see name.

UUID

Remove the <uuid> line, a new UUID will be generated.

Memory

Adjust the amount of memory, specified as <memory>. Remove the <currentMemory> line.

Virtual CPU Count

Adjust the number of virtual CPUs allocated to this virtual machine, using <vcpu> element. Use an "auto" placement:

<vcpu placement='auto'>2</vcpu>

CDROM

Remove the "cdrom" disk(s), if it is not going to be used.

Storage

Edit the <disk> definitions and adjust the names for the new storage volumes that will be provisioned for the new virtual machine. The virtual machine will definitely need a virtual machine image, stored in (usually) qcow2 format, and optionally other block storage devices in raw format. Conventionally, the virtual machine image is stored in the main storage pool and it is named based on the name of the VM. For more naming conventions, see Linux Virtualization Naming Conventions.

<disk type='file' device='disk'>
  <driver name='qemu' type='qcow2'/>
  <source file='/main-storage-pool/new-vm-name.qcow2'/>
  <target dev='vda' bus='virtio'/>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/>
</disk>

<disk type='file' device='disk'>
  <driver name='qemu' type='raw'/>
  <source file='/main-storage-pool/'new-vm-name-docker.raw'/>
  <target dev='vdb' bus='virtio'/>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x08' function='0x0'/>
</disk>

If an additional disk is needed, it can be created externally as a storage volume, and then attached to the template, after the template was shut down:

virsh attach-disk <template-name> /main-storage-pool/'<template-name>-docker.raw vdb --config

Network

The two most common situations are configuring the guest with a network interface that connects to a virtual network or with a network interface that attached directly to one of the virtualization host physical interfaces. The background behind these choices is covered extensively in the Virtual Networking Concepts document.

If we want the guest to connect to a virtual network, the guest network interface should be configured as described here: Guest Configuration for Virtual Network.

If we want the guest to stand up a network interface that connects directly to a virtualization host's physical network interface with a macvtap driver, the guest network interface should be configured as described here: Guest Configuration for Direct Attachment to Physical Device with macvtap Driver.

Locate the network interface and replace the value of the mac address with a randomly generated value.

It is important to replace the MAC address with a unique value per virtualization host, otherwise conflicts will occur.

A shell script that generate random MAC addresses is available here: bash script that generates a random MAC address.

<interface type='network'>
    <mac address='52:54:00:79:03:0c'/>
    <source network='default'/>
    <model type='virtio'/>
    <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/>
</interface>

See <interface>.

Provision the Storage

For each disk the virtual machine relies on, starting with the virtual machine image storage, clone the disks following the appropriate procedure. If the storage was originally empty, or we want to start with an empty storage, we could provision a new disk, as described below. Note that this does not apply to the virtual machine image, we need the content of the virtual machine image to clone the virtual machine:

• Clone a qcow2 virtual machine image
• Clone a raw volume
• Create a new raw block storage volume (a new filesystem will need to be built on this one later).

When creating logical volumes, follow the storage volume naming conventions. For VirtualBox, the procedure of creating and installing a new virtual disk is available here.

Create the Clone

virsh define .../vm-definition-repository/target_vm.xml

virsh define /root/environments/ocp36/ocp36.generic-template.xml

Edit Virtual Machine Image before the First Boot

This step is optional, but the virtual machine image can be edited off-line before boot with virt-edit, and configuration details such as the IP address(es), can be configured, so the virtual machine starts without IP address conflicts.

 virt-edit -a /main-storage-pool/appproxy.qcow2 /etc/sysconfig/network-scripts/ifcfg-eth0
 virt-edit -a /main-storage-pool/appproxy.qcow2 /etc/hosts
 ...

Boot the Clone and Finalize the Configuration

Boot the clone:

virsh start --console new-virtual-machine-name

virsh start --console ocp36.generic-template

Connect to it with virsh console, or via SSH using the IP address of the template, and finalize the configuration:

Reconfigure Linux VM Guest Image