I bought a new PC and installed KVM, so I summarized the procedure.

[1. Linux installation](#1-Linux のインストール)

First, install Linux. This time I installed Ubuntu 20.04.

[2. Install KVM](#2-KVM のインストール)

Install the VM-related packages on the KVM host.

$ sudo apt update
$ sudo apt install -y qemu qemu-kvm libvirt-daemon libvirt-clients bridge-utils virt-manager
$ sudo reboot

Then add the user you are using to the libvirt and kvm groups. Not required if it has already been added.

$ sudo adduser `id -un` libvirt
Adding user '<username>' to group 'libvirt' ...
$ sudo adduser `id -un` kvm
Adding user '<username>' to group 'kvm' ...

You may also want to install tools such as virt-top and virt-df. Install Python, jq, etc. if necessary.

3. Disk Configuration

For example, if you want to use a disk other than the system disk for the qcow2 image, format the disk. First of all, the parted used for partition creation was helpful here.

4. Check Network

By default, a virtual bridge called virbr0 is created during KVM installation.

$ ip addr show virbr0
4: virbr0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN group default qlen 1000
    link/ether 52:54:00:5e:9e:a2 brd ff:ff:ff:ff:ff:ff
    inet 192.168.122.1/24 brd 192.168.122.255 scope global virbr0
       valid_lft forever preferred_lft forever
$ brctl show
bridge name     bridge id               STP enabled     interfaces
virbr0          8000.5254005e9ea2       yes             virbr0-nic
$ virsh net-list
 Name      State    Autostart   Persistent
--------------------------------------------
 default   active   yes         yes

This bridge can be used as NAT to allow virtual machines to access the outside network. If you don't use it, you can delete it. Of course, you don't have to erase it. If you want to erase it, this is the area.

$ virsh net-destroy default
$ virsh net-undefine default

If you look at here, it says that the following sysctl parameters should be set to 0.

net.bridge.bridge-nf-call-ip6tables=0
net.bridge.bridge-nf-call-iptables=0
net.bridge.bridge-nf-call-arptables=0

This is probably the case when a module called br_netfilter is loaded. I don't think it's needed in Ubuntu 20.04 as it wasn't loaded by default. Maybe w

5. Network Settings

If you want to communicate from the outside to the IP address assigned to the virtual machine, create a bridge on Linux. When using netplan on Ubuntu 20.04, it looks like this.

$ cat /etc/netplan/60-eno1-config.yaml
network:
    version: 2
    ethernets:
        eno1:
            dhcp4: false
            dhcp6: false
    bridges:
        br0:
            interfaces:
                - eno1
            dhcp4: false
            dhcp6: false
            addresses:
                - 10.0.0.249/24
            gateway4: 10.0.0.1
            nameservers:
                addresses:
                    - 1.1.1.1
                    - 1.0.0.1
            parameters:
                forward-delay: 0
                stp: false
            optional: true

You can explicitly register the above with KVM, but the KVM guest will be visible when you create the bridge on the KVM host. What to do with STP depends on the text, but many are invalid. It doesn't seem to fall even if it is effective.

Reference: I investigated the virtual network used by KVM on CentOS

This is probably because it is used by default on CentOS, but there are many examples using Network Manager. You can also go around here.

You can check the created bridge with the brctl command.

$ brctl show
bridge name     bridge id               STP enabled     interfaces
br0             8000.a0369fa9f098       no              enp4s0f0

By assigning the bridge created above to the KVM guest and assigning an IP address, the corresponding IP can be accessed from outside the KVM environment.

6. Create an internal network

Create a network for use inside KVM. You can also create it with virt-manager, but here I will create it from an xml file.

First, create an xml file. I think the following is probably the simplest description. You can also set the IP address, DHCP, etc. here.

$ cat virbr1.xml
<network>
  <name>virbr1</name>
  <bridge name='virbr1' stp='on' delay='0' />
</network>

Reference: Procedure for building a virtual environment using KVM (Kernel-based Virtual Machine) (Procedure for installing KVM and creating a virtual network)

Create a bridge in KVM with the following command so that it is automatically configured at startup.

$ virsh net-define ./virbr1.xml
Network virbr1 defined from ./virbr1.xml

$ virsh net-autostart virbr1
Network virbr1 marked as autostarted

$ virsh net-start virbr1
Network virbr1 started

You can check if it was created normally with the following command.

$ brctl show
bridge name     bridge id               STP enabled     interfaces
virbr1          8000.525400a99a82       yes             virbr1-nic
$ virsh net-list
 Name      State    Autostart   Persistent
--------------------------------------------
 default   active   yes         yes
 virbr1    active   yes         yes

The actual config file is here. The uuid and mac address are automatically assigned. Once registered, it can be edited with the virsh net-edit command.

$ sudo cat /etc/libvirt/qemu/networks/virbr1.xml
<!--
WARNING: THIS IS AN AUTO-GENERATED FILE. CHANGES TO IT ARE LIKELY TO BE
OVERWRITTEN AND LOST. Changes to this xml configuration should be made using:
  virsh net-edit virbr1
or other application using the libvirt API.
-->

<network>
  <name>virbr1</name>
  <uuid>379d3199-365e-4c99-b331-77d50ed7b2ac</uuid>
  <bridge name='virbr1' stp='on' delay='0'/>
  <mac address='52:54:00:a9:9a:82'/>
</network>

[7. Run virsh remotely](#7-virsh をリモートで実行する)

You may not want to log in directly to the KVM host, such as with ssh. In virsh and virt-manager he configures KVM, but it can also be run remotely.

To run virsh remotely, first install libvirt-clients on your client.

$ sudo apt install libvirt-clients

Then execute the following command.

$ virsh -c qemu+ssh://[email protected]/system list
 Id   Name   State
----------------------
 1    c20h   running

In the above case, since you are connecting with ssh, you need to make settings such as ssh public key authentication. Also, it may be convenient to write the following in .bash_aliases etc.

$ cat .bash_aliases | grep vir
alias virsh='virsh -c qemu+ssh://[email protected]/system'
$ virsh list
 Id   Name   State
----------------------
 1    c20h   running

virt-manager can be configured from the GUI File> Add Connection. Both virsh and virt-manager are convenient because they can be executed on Linux on Windows Subsystem for Linux (WSL).

See also: SSH Setup (https://wiki.libvirt.org/page/SSHSetup)

[8. Migrate KVM guests](#8-KVM ゲストを移行する)

At least two files must be copied from the old KVM host to the new KVM host when migrating KVM guests. The KVM guest image body (qcow2) and configuration file (xml) file.

The KVM guest image has a qcow2 extension, and there may be a better way, but you can find the location with the virsh dumpxml command. When copying the qcow2 file, I think it is better to shut down her KVM guest in advance.

$ virsh dumpxml c20h | grep .qcow2
      <driver name='qemu' type='qcow2'/>
      <source file='/mnt/datastore1/images/c20h.qcow2'/>

To get the KVM guest configuration file (xml), use the virsh dumpxml command as well.

$ virsh dumpxml c20h > c20h.xml

To register a KVM guest to migrate to a new KVM host, run the following command.

$ virsh define c20h.xml

The first thing to consider before running the above command is the location of the qcow2 file. If you place the qcow2 file in the same directory, you do not need to change it, but if it is different, pre-edit the xml file with a text editor.

<source file='/mnt/datastore1/images/c20h.qcow2'/>

Also, there may be a description of the cpu type in the xml file. If the CPU type is different, change this description.

  <cpu mode='custom' match='exact' check='partial'>
    <model fallback='allow'>Haswell-noTSX-IBRS</model>
  </cpu>

I think it's a good idea to simply change it so that it inherits the CPU information of the KVM host.

  <cpu mode='host-model' check='partial'/>

The above settings can also be changed with virtsh or virt-manage after using the virsh define command. It may be easy and good to change it with virt-manager.

Executing the virsh define command creates a configuration file under/etc/libvirt/qemu.

$ ls -al /etc/libvirt/qemu/c20h.xml
-rw------- 1 root root 4663 Jan 10 18:10 /etc/libvirt/qemu/c20h.xml

See also: Linux Notes Moving Virtual Machines Between KVM Virtualization Hosts (https://oplern.hatenablog.com/entry/2016/11/06/212257)

After migration, the KVM guest can be started by connecting to the console with virt-manager or virt-viewer and changing the settings.

You can mount and edit KVM guest disks by installing libguestfs-tools. This way you can change the settings before starting the KVM guest.

$ sudo guestmount -d c20h -i /mnt/c20h/
$ sudo ls /mnt/c20h/etc/netplan
50-cloud-init.yaml
$ sudo vi /mnt/c20h/etc/netplan
$ sudo vi /mnt/c20h/etc/netplan/50-cloud-init.yaml

See also: How to mount VM virtual disk on KVM hypervisor with Libguestfs Tools (https://computingforgeeks.com/how-to-mount-vm-virtual-disk-on-kvm-hypervisor/)

[LINUX] Install and configure KVM