When thinking about virtualization, everybody immediately thinks about VMWare. And it must be said, the product they offer is very decent but also comes with a “decent” price. As an alternative, it’s worth looking into KVM for your virtualization. As with the VMWare product range, KVM offers full virtualization and it can compete with VMWare regarding stability and performance.
Update for CentOS 8 and RHEL 8
I have completely rewritten this article for CentOS 8 and RHEL 8 in the following post: Install and use CentOS 8 or RHEL 8 as KVM virtualization host
If you’re looking for the guide for el7, please ignore this and continue reading :)
To prevent things getting confused I would first like to clear out some terminology used for virtualization. For somebody working on a daily basis in virtual environments, these might be clear but can be rather confusing for others.
Host: the machine that hosts other system, KVM will be installed on this machine
Guest: the system running on the host, also referred to as VM, Virtual Machine or domain.
Hypervisor: the piece of software that enables virtualization on the host. For example: KVM, ESXi, Xen, …
This post has been completely reviewed and updated where needed on 08/2017 to make sure everything is still working and is correct.
Part1: KVM installation and preparation
KVM hypervisor and VM-extensions
As mentioned earlier, KVM offers, as VMWare, full virtualization. This means that a full system, which looks like a real physical system to the guest-OS, will be offered. Besides full virtualization, there is also such a thing as paravirtualization, as Xen can offer. Paravirtualization gives you higher performance but needs a modified guest-OS and is basically limited to *nix-systems. Full virtualization enables you to run unmodified guest-systems and thus also most proprietary systems as Windows . In order to be able to use full virtualization, you either need some virtualization-extensions on your CPU or use emulation.
First thing to do is to check if the host-machine supports VM-extensions. On the x86 platofrom, those are either AMD-V or Intel’s VT-X. In order to check if the installed CPU’s support those extensions, we need to check if the vmx (for VT-X) or svm (for AMD-V) flag exists in the cpuinfo-output:
[jensd@kvmhost ~]$ egrep -c '(vmx|svm)' /proc/cpuinfo 2
When the output is 0, meaning that neither vmx or svm is found in the flags, it probably means that your CPU doesn’t support those extensions and there is little you can do. When the extensions are listed, be sure to check if they are enabled in the systems BIOS since that would cause problems later on. In case your CPU doesn’t support VM-extensions, you are limited to QEMU-emulation in combination with KVM, which delivers a much worse performance in comparison. For this tutorial, I’ll assume that the VM-extensions are supported and enabled in the BIOS of the host-system.
The first step in the KVM installation is installing the necessary packages. Package virt-manager, xauth and dejavu-lgc-sans-fonts are also needed if you want to manage KVM with the graphical interface in combination with X11 forwarding. (for more information, check this previous post about X11 forwarding)
To install the required packages
amp;nbsp;sudo yum install kvm virt-manager libvirt virt-install qemu-kvm xauth dejavu-lgc-sans-fonts -y ... Complete!
For the networking part, our KVM-host will act as a router for its guests and we will need to create a bridge interface to allow the guest to communicate out of the host. Guests will use the bridge on the host to connect to the real network. To allow such type of setup it’s needed to allow ip forwarding in the kernel parameters.
[jensd@kvmhost ~]$ echo "net.ipv4.ip_forward = 1"|sudo tee /etc/sysctl.d/99-ipforward.conf net.ipv4.ip_forward = 1 [jensd@kvmhost ~]$ sudo sysctl -p /etc/sysctl.d/99-ipforward.conf net.ipv4.ip_forward = 1
After allowing the host to do ip forwarding, we need to change the network configuration. Basically we will keep our original physical interface as it is but will assign its IP-address to the brige. In the example host-machine there is one real interface called eno16777736 and the script in /etc/sysconfig/network-scripts/ifcfg-eno16777736 looks like this:
DEVICE="eno16777736" ONBOOT=yes IPADDR="192.168.202.111" NETMASK="255.255.255.0" GATEWAY="192.168.202.2" HWADDR="00:0c:29:32:d0:4c" DNS1="192.168.202.2"
The first thing to change here, is to comment out everything that is IP-related and tell the interface which interface will be the bridge. Resulting in /etc/sysconfig/network-scripts/ifcfg-eno16777736 to look like this:
DEVICE="eno16777736" ONBOOT=yes #IPADDR="192.168.202.111" #NETMASK="255.255.255.0" #GATEWAY="192.168.202.2" HWADDR="00:0c:29:32:d0:4c" #DNS1="192.168.202.2" BRIDGE=virbr0
Next, we can create the config-script for the bridge interface virbr0 in /etc/sysconfig/network-scripts/ifcfg-virbr0. Most details can be copied from the original script for eno16777736:
DEVICE="virbr0" TYPE=BRIDGE ONBOOT=yes BOOTPROTO=static IPADDR="192.168.202.111" NETMASK="255.255.255.0" GATEWAY="192.168.202.2" DNS1="192.168.202.2"
Finish and check the KVM installation
Basically all components are now ok but before KVM can be used it’s a good idea to perform a reboot in order to load the kvm-modules and to relaod the new network settings.
After the reboot, we should check if the necessary kernel modules are loaded, which means that KVM successfully can handle the VM-extensions of our CPU:
[jensd@kvmhost ~]$ lsmod|grep kvm kvm_intel 138567 0 kvm 441119 1 kvm_intel
Check if the bridge is installed and in an up-state:
[jensd@kvmhost ~]$ ip a show virbr0 3: virbr0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP link/ether 00:0c:29:32:d0:4c brd ff:ff:ff:ff:ff:ff inet 192.168.202.111/24 brd 192.168.202.255 scope global virbr0 valid_lft forever preferred_lft forever inet6 fe80::20c:29ff:fe32:d04c/64 scope link valid_lft forever preferred_lft forever
Last thing to check is if we can connect to KVM by asking for a simple list of systems:
[jensd@kvmhost ~]$ sudo virsh -c qemu:///system list Id Name State ----------------------------------------------------
If it returns something else, then you should go trough the earlier steps to check where something went wrong.
Part 2: Using KVM with the CLI
After completing the KVM installation, it’s time to start using the host. First thing we need to do is to create a new domain or VM.
Adding a new VM
To create a new virtual machine using the CLI, we need to know which template we will use to install the system. To get a list of templates that are known in our KVM installation, you can do the following:
[jensd@cen7 ~]$ osinfo-query os Short ID | Name | Version | ID ----------------------+----------------------------------------------------+----------+----------------------------------------- altlinux1.0 | Mandrake RE Spring 2001 | 1.0 | http://altlinux.org/altlinux/1.0 altlinux2.0 | ALT Linux 2.0 | 2.0 | http://altlinux.org/altlinux/2.0 altlinux2.2 | ALT Linux 2.2 | 2.2 | http://altlinux.org/altlinux/2.2 altlinux2.4 | ALT Linux 2.4 | 2.4 | http://altlinux.org/altlinux/2.4 altlinux3.0 | ALT Linux 3.0 | 3.0 | http://altlinux.org/altlinux/3.0 altlinux4.0 | ALT Linux 4.0 | 4.0 | http://altlinux.org/altlinux/4.0 altlinux4.1 | ALT Linux 4.1 | 4.1 | http://altlinux.org/altlinux/4.1 altlinux5.0 | ALT Linux 5.0 | 5.0 | http://altlinux.org/altlinux/5.0 altlinux6.0 | ALT Linux 6.0 | 6.0 | http://altlinux.org/altlinux/6.0 altlinux7.0 | ALT Linux 7.0 | 7.0 | http://altlinux.org/altlinux/7.0 centos6.0 | CentOS 6.0 | 6.0 | http://centos.org/centos/6.0 centos6.1 | CentOS 6.1 | 6.1 | http://centos.org/centos/6.1 centos6.2 | CentOS 6.2 | 6.2 | http://centos.org/centos/6.2 centos6.3 | CentOS 6.3 | 6.3 | http://centos.org/centos/6.3 centos6.4 | CentOS 6.4 | 6.4 | http://centos.org/centos/6.4 centos6.5 | CentOS 6.5 | 6.5 | http://centos.org/centos/6.5 centos6.6 | CentOS 6.6 | 6.6 | http://centos.org/centos/6.6 centos6.7 | CentOS 6.7 | 6.7 | http://centos.org/centos/6.7 centos7.0 | CentOS 7.0 | 7.0 | http://centos.org/centos/7.0 debian1.1 | Debian Buzz | 1.1 | http://debian.org/debian/1.1 debian1.2 | Debian Rex | 1.2 | http://debian.org/debian/1.2 debian1.3 | Debian Bo | 1.3 | http://debian.org/debian/1.3 debian2.0 | Debian Hamm | 2.0 | http://debian.org/debian/2.0 debian2.1 | Debian Slink | 2.1 | http://debian.org/debian/2.1 debian2.2 | Debian Potato | 2.2 | http://debian.org/debian/2.2 debian3 | Debian Woody | 3 | http://debian.org/debian/3 debian3.1 | Debian Sarge | 3.1 | http://debian.org/debian/3.1 debian4 | Debian Etch | 4 | http://debian.org/debian/4 debian5 | Debian Lenny | 5 | http://debian.org/debian/5 debian6 | Debian Squeeze | 6 | http://debian.org/debian/6 debian7 | Debian Wheezy | 7 | http://debian.org/debian/7 debian8 | Debian Jessie | 8 | http://debian.org/debian/8 fedora-unknown | Fedora | unknown | http://fedoraproject.org/fedora/unknown fedora1 | Fedora Core 1 | 1 | http://fedoraproject.org/fedora/1 fedora10 | Fedora 10 | 10 | http://fedoraproject.org/fedora/10 fedora11 | Fedora 11 | 11 | http://fedoraproject.org/fedora/11 fedora12 | Fedora 12 | 12 | http://fedoraproject.org/fedora/12 fedora13 | Fedora 13 | 13 | http://fedoraproject.org/fedora/13 fedora14 | Fedora 14 | 14 | http://fedoraproject.org/fedora/14 fedora15 | Fedora 15 | 15 | http://fedoraproject.org/fedora/15 ...
Virtual disk images for the KVM-guests can be placed in /var/lib/libvirt by default. In case you prefer to use another location to store the disk images, SELinux will, by default, prevent access and the security context of that location needs to be changed in order to use it for KVM. To change the SELinux context when storing the images in another location (/vm for example):
[jensd@kvmhost ~]$ sudo yum install policycoreutils-python -y ... Complete! [jensd@kvmhost ~]$ sudo mkdir /vm [jensd@kvmhost ~]$ sudo semanage fcontext -a -t virt_image_t "/vm(/.*)?" [jensd@kvmhost ~]$ sudo restorecon -R /vm
Now, to add a new VM, we can use virt-install.
Example to add a windows-guest:
[jensd@kvmhost ~]$ sudo virt-install --connect qemu:///system -n vmwin7 -r 1024 --vcpus=2 --disk path=/var/lib/libvirt/images/vmwin7.img,size=10 --graphics vnc,listen=0.0.0.0 --noautoconsole --os-type windows --os-variant win7 --accelerate --network=bridge:virbr0 --hvm --cdrom /var/X17-59186.iso Starting install... Allocating 'vmwin7.img' | 10 GB 00:00:00 Creating domain... | 0 B 00:00:00 Domain installation still in progress. Waiting for installation to complete.
Explanation of the arguments that were given to virt-install:
- –connect qemu:///system : connect to KVM on the local system, we could also connect to another KVM-host and define our new VM there
- -n vmwin7 : name of the new VM: vmwin7
- -r 1024 : amount of memory for the VM: 1GB
- –vcpus=2 : amount of virtual CPU’s for the VM: 2
- –disk path=/var/lib/libvirt/images/vmwin7.img,size=10 : where to store the virtual disk image of the VM and the size: 10GB
- –graphics vnc,listen=0.0.0.0 : how to display the VM’s console: via VNC accessible from outside
- –noautoconsole : do not automatically connect to the console
- –os-type windows –os-variant win7 : type of guest OS (from the list given above)
- –accelerate : use KVM HW-acceleration
- –network=bridge:virbr0 : network bridge to use
- –hvm : full virtualisation
- –cdrom /var/X17-59186.iso : location of the installation ISO
After launching the above command, you should be able to connect with VNC to the host and get on the console-display of the VM. The console displays what would normally, on a physical machine, appear on the attached monitor.
By default, VNC will use the first available screen on port 5900. To be sure which screen is used, we can use virsh to show the attached console-screens for VNC:
[jensd@kvmhost ~]$ sudo virsh vncdisplay vmwin7 :0
:0 means the first screen and real port 5900 as you can also see when checking with netstat which ports are currently listening:
[jensd@kvmhost ~]$ netstat -tln|grep :59 tcp 0 0 0.0.0.0:5900 0.0.0.0:* LISTEN
Now, connect to the KVM-host with a VNC viewer. I’m using TightVNC but every VNC viewer should do:
From this point, we can complete the windows installation as if it would be a normal physical system:
After completing the installation with VNC, we end up with a Windows-VM that is running on our KVM-host:
As for the networking part, we use the earlier created bridge (virbr0) to do NAT. This means that the KVM-host NAT’s all our connections to the real network connected to the KVM-host. If DHCP is active on that network, it can be used in the VM. Otherwise you will have to configure a static IP in the same subnet.
Example to add a Linux-guest:
To add a Linux guest, next to the already added Windows-guest is quite similar:
[jensd@kvmhost ~]$ sudo virt-install --connect qemu:///system -n vmdeb7 -r 512 --vcpus=1 --disk path=/var/lib/libvirt/images/vmdeb7.img,size=2 --graphics vnc,listen=0.0.0.0 --noautoconsole --os-type linux --os-variant debianwheezy --accelerate --network=bridge:virbr0 --hvm --cdrom /tmp/debian-7.5.0-amd64-netinst.iso Starting install... Allocating 'vmdeb7.img' | 2.0 GB 00:00:00 Creating domain... | 0 B 00:00:02 Domain installation still in progress. You can reconnect to the console to complete the installation process.
As with the Windows-VM, after launching this command, you should be able to connect with VNC to the host and get on the console of the VM to complete the Debian installation.
To know which VNC-display number (and port) is used for a certain VM, the same command as used earlier should do:
[jensd@kvmhost ~]$ sudo virsh vncdisplay vmdeb7 :1
Above command gives :1 as result, meaning that the guest vmdeb7 can be contacted with VNC on port 5901:
After finishing the installation, we end up with a Linux guest running on top of our KVM-host. Which Linux distro we are using doesn’t matter since we’re doing full virtualization.
root@deb:~# uname -a Linux deb 3.2.0-4-amd64 #1 SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux root@deb:~# cat /proc/cpuinfo |grep model model : 13 model name : QEMU Virtual CPU version 1.5.3
Considering network, the same as with the Windows VM applies here. Our connections are NATted trough the KVM-host and we can use the DHCP-server of our real network.
root@deb:~# ping -c1 22.214.171.124 PING 126.96.36.199 (188.8.131.52) 56(84) bytes of data. 64 bytes from 184.108.40.206: icmp_req=1 ttl=128 time=23.8 ms --- 220.127.116.11 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 23.855/23.855/23.855/0.000 ms root@deb:~# ip a show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether 52:54:00:33:65:75 brd ff:ff:ff:ff:ff:ff inet 192.168.202.140/24 brd 192.168.202.255 scope global eth0 inet6 fe80::5054:ff:fe33:6575/64 scope link valid_lft forever preferred_lft forever
More KVM actions
Besides creating VM’s, it’s a good thing to know some basic operations regarding VM-managent.
List the active virtual machines:
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system list Id Name State ---------------------------------------------------- 7 vmwin7 running 8 vmdeb7 running
Get more information about a guest:
[jensd@kvmhost ~]$ sudo virsh dominfo vmwin7 Id: 7 Name: vmwin7 UUID: f913c6fa-b597-437d-b6f5-797314e34847 OS Type: hvm State: running CPU(s): 2 CPU time: 20955.1s Max memory: 1048576 KiB Used memory: 1048576 KiB Persistent: yes Autostart: disable Managed save: no Security model: selinux Security DOI: 0 Security label: system_u:system_r:svirt_t:s0:c638,c926 (enforcing)
Stop a running guest:
To stop a running VM in a clean way (as you would press the power button to start the shutdown sequence):
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system shutdown vmdeb7 Domain vmdeb7 is being shutdown
This triggers a normal, clean, shutdown on the guest:
root@deb:~# Broadcast message from root@deb (Wed Aug 27 09:09:16 2014): Power button pressed The system is going down for system halt NOW!
To force stop a running VM that doesn’t want to shutdown in a clean way:
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system destroy vmdeb7 Domain vmdeb7 destroyed
Start a guest:
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system start vmdeb7 Domain vmdeb7 started
Delete a guest:
First we need to make sure that the guest is stopped before it can be deleted. In case you don’t want the virtual disk image anymore either, you’ll have to delete it manually after undefining the guest.
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system destroy vmcen6 Domain vmcen6 destroyed [jensd@kvmhost ~]$ sudo virsh --connect qemu:///system undefine vmcen6 Domain vmcen6 has been undefined [jensd@kvmhost ~]$ sudo rm -f /var/lib/libvirt/images/vmcen6.img
After removing a disk-image, it’s a good thing to refresh the storage pool of KVM:
[jensd@kvmhost ~]$ sudo virsh pool-refresh default Pool default refreshed
Automatically let a guest start when the host starts
When rebooting your host, you probably want some or all the guests that are defined on that host to start at the same time. By default, the guest are not automatically started.
[jensd@kvmhost ~]$ sudo virsh --connect qemu:///system autostart vmdeb7 Domain vmdeb7 marked as autostarted [jensd@kvmhost ~]$ sudo virsh --connect qemu:///system dominfo vmdeb7|grep Auto Autostart: enable
Part 3: Using KVM with the virt-manager GUI
Starting the GUI
Managing KVM with the CLI is not so difficult and it can be very handy to script certain day-to-day tasks. Sometimes, you just need to keep an overview and require a little more user-friendliness. For that, you can use virt-manager, which is a graphical interface for libvirt and is mainly built for KVM. When you want to manage your guest with virt-manager, you can either do it on the host itself, by starting an X-server locally or use X11 forwarding on a headless server (more information here).
Make sure that you have enough permissions to use virt-manager and simply execute virt-manager from the command line:
[root@kvmhost ~]# virt-manager
If all goes well, you should be presented with the virt-manager GUI:
From the initial start-up screen, you can immediately see a list of configured guests on this host and take actions on them like: Run, Pause, Shutdown, Reboot, Force off,…
When selecting a guest, you can also click on Open to display the console as we did earlier using VNC:
Other possibilities using the virt-manager interface:
Connect to another host-system to manage the VM’s running there, using File -> Add connection (like using the –connect on the CLI)
Migrate a VM to another KVM-host: right click on the VM and choose Migrate…
Clone a VM to the same or another KVM-host: right click on the VM and choose Clone…
As you can see, the virt-manager interface is not very complicated and most of the basic tasks don’t need any explanation.
After completing all of the above steps, basic installation and using KVM shouldn’t have any secrets anymore for you. The next thing to do is experiment and test a little more with KVM and hopefully start to use it in your production environment.