Difference between revisions of "Libvirt"

From ArchWiki
Jump to: navigation, search
(Host configuration: screw rc.local, use POST_UP instead)
m (Enable KVM acceleration for QEMU: adjust grep command)
Line 100: Line 100:
Running virtual machines with the usual [[QEMU]] emulation (i.e. without KVM)), will be '''painfully slow'''. You definitely want to enable KVM support if your CPU supports it. To find out, run the following command:
Running virtual machines with the usual [[QEMU]] emulation (i.e. without KVM)), will be '''painfully slow'''. You definitely want to enable KVM support if your CPU supports it. To find out, run the following command:
  grep -E "(vmx|svm)" --color=always /proc/cpuinfo
  egrep --color "vmx|svm" /proc/cpuinfo
If that command generates output, then your CPU supports hardware acceleration via KVM; if that command does ''not'' generate output, then you ''cannot use KVM''.
If that command generates output, then your CPU supports hardware acceleration via KVM; if that command does ''not'' generate output, then you ''cannot use KVM''.

Revision as of 15:55, 28 November 2012

Summary help replacing me
This article does not try to cover everything about libvirt, just the things that were not intuitive at first or not well documented.

libvirt is a virtualization API and a daemon for managing virtual machines (VMs) -- remote or locally, using multiple virtualization back-ends (QEMU/KVM, VirtualBox, Xen, etc).


For servers you need the following packages from the official Arch Linux repositories:

For GUI management tools, you also need all of the following from the official Arch Linux repositories:

Building libvirt for Xen

The PKGBUILD for both libvirt-gitAUR in the AUR and libvirt in the official repositories currently disables Xen support with the --without-xen flag during the make process. If you want to use libvirt for managing Xen, you will need to grab the whole file set to enable Xen support and build your own libvirt package using the Arch Build System. Furthermore, you need to make sure you have libxenctrlAUR installed. If xenAUR is installed, you don't need to install libxenctrlAUR.

The alternative XenAPI driver is lacking a package at the moment? (2010-05-23, friesoft)


Run daemon

Start the libvirtd daemon and add libvirtd to your DAEMONS array so it starts automatically on boot.

It seems like you have to start dbus and avahi-daemon before starting libvirtd.

Note: The Avahi daemon is used for local discovery of libvirt hosts via multicast-DNS. To disable this functionality, set mdns_adv = 0 in /etc/libvirt/libvirtd.conf.

PolicyKit authentication

To allow a non-root user to manage virtual machines, you need to create the following file:

PolicyKit authorization (version < 0.107)

 [Allow a user to manage virtual machines]
 Identity=unix-user:<replace with your username>

PolicyKit authorization (version >= 0.107)

polkit.addRule(function(action, subject) {
    if (action.id == "org.libvirt.unix.manage" && subject.user == "<replace with user name>") {
        return polkit.Result.YES;

Try to log in again if it does not work right away.

Alternatively, you can grant only the monitoring rights with org.libvirt.unix.monitor

For more information, see the libvirt wiki.

Unix file-based permissions

Note: This is an alternative to PolicyKit authentication.

If you wish to use Unix file-based permissions to allow some non-root users to use libvirt, you can modify the configuration files.

First, you will need to create the libvirt group and add any users you want to have access to libvirt to that group.

# groupadd libvirt
# gpasswd -a [username] libvirt

Any users that are currently logged in will need to log out and log back in to update their groups. Alternatively, the user can use the following command in the shell they will be launching libvirt from to update the group:

$ newgrp libvirt
Note: The line to uncomment for unixperms has been removed from PKGBUILD into this patch

Then you can either enable permissions-based access by uncommenting the following line in the PKGBUILD for libvirt before running makepkg -s:

# patch -Np1 -i "$srcdir"/unixperms.patch

Alternatively, you can make the changes to your permissions and configuration files by hand. Uncomment the following lines in /etc/libvirt/libvirtd.conf (they are not all in the same location in the file):

 #unix_sock_group = "libvirt"
 #unix_sock_ro_perms = "0777"
 #unix_sock_rw_perms = "0770"
 #auth_unix_ro = "none"
 #auth_unix_rw = "none"
Note: You may also wish to change unix_sock_ro_perms from 0777 to 0770 to disallow read-only access to people who are not members of the libvirt group.

Enable KVM acceleration for QEMU

Note: KVM will conflict with VirtualBox. You cannot use KVM and VirtualBox at the same time.

Running virtual machines with the usual QEMU emulation (i.e. without KVM)), will be painfully slow. You definitely want to enable KVM support if your CPU supports it. To find out, run the following command:

egrep --color "vmx|svm" /proc/cpuinfo

If that command generates output, then your CPU supports hardware acceleration via KVM; if that command does not generate output, then you cannot use KVM.

To enable KVM, you need to load the kvm-amd or kvm-intel kernel module depending on your CPU. In recent Linux kernels, these modules are loaded automatically.

If KVM is not working, you will find the following message in your /var/log/libvirt/qemu/VIRTNAME.log:

 Could not initialize KVM, will disable KVM support

More info is available from the official KVM FAQ

Stopping / resuming guest at host shutdown / startup

Running guests may be suspended (or shutdown) at host shutdown automatically using the libvirt-guests service. On the other hand, at host startup, this same daemon will resume (startup) the suspended (shutdown) guests automatically. Check /etc/conf.d/libvirtd-guests for libvirt-guests options.


Installing a new VM

To create a new VM, you need some sort of installation media, which is usually a standard .iso file. Copy it to the /var/lib/libvirt/images/ directory (alternatively, you can create a new storage pool directory in virt-manager and copy it there).

Note: SELinux requires that virtual machines be stored in /var/lib/libvirt/images/ by default. If you use SELinux and are having issues with virtual machines, ensure that your VMs are in that directory or ensure that you have added the correct labeling to the non-default directory that you used.

Then run virt-manager, connect to the server, right click on the connection and choose New. Choose a name, and select Local install media. Just continue with the wizard.

On the 4th step, you may want to uncheck Allocate entire disk now -- this way you will save space when your VM is not using all of its disk. However, this can cause increased fragmentation of the disk, and you must pay attention to the total available disk space on the VM host because it is much easier to over-allocate disk space to VMs.

On the 5th step, open Advanced options and make sure that Virt Type is set to kvm. If the kvm choice is not available, see section Enable KVM acceleration for QEMU above.

Creating a storage pool in virt-manager

First, connect to an existing server. Once you are there, right click and choose Details. Go to Storage and press the + icon at the lower left. Then just follow the wizard. :)

Using VirtualBox with virt-manager

Note: VirtualBox support in libvirt is not quite stable yet and may cause your libvirtd to crash. Usually this is harmless and everything will be back once you restart the daemon.

virt-manager does not let you to add any VirtualBox connections from the GUI. However, you can launch it from the command line:

virt-manager -c vbox:///system

Or if you want to manage a remote system over SSH:

virt-manager -c vbox+ssh://username@host/system

Remote access to libvirt

Using unencrypted TCP/IP socket (most simple, least secure)

Warning: This should only be used for testing or use over a secure, private, and trusted network.

Edit /etc/libvirt/libvirtd.conf:

listen_tls = 0
listen_tcp = 1
Warning: We do not enable SASL here, so all TCP traffic is cleartext! For real world use, always enable SASL.

It is also necessary to start the server in listening mode by editing /etc/conf.d/libvirtd


Using SSH

The openbsd-netcat package is needed for remote management over SSH.

To connect to the remote system using virsh:

$ virsh -c qemu+ssh://username@host/IP address/system

If something goes wrong, you can get some logs using:

$ LIBVIRT_DEBUG=1 virsh -c qemu+ssh://username@host/IP address/system

To display the graphical console for a virtual machine:

$ virt-viewer --connect qemu+ssh://username@host/IP address/system myvirtualmachine

To display the virtual machine desktop management tool:

$ virt-manager -c qemu+ssh://username@host/IP address/system
Note: If you are having problems connecting to a remote RHEL server (or anything other than Arch, really), try the two workarounds mentioned in FS#30748 and FS#22068.

Using Python

The libvirt package comes with a python2 API in /usr/lib/python2.7/site-packages/libvirt.py

General examples are given in /usr/share/doc/libvirt-python-your_libvirt_version/examples/

Unofficial example using qemu and openssh:

#! /usr/bin/env python2
# -*- coding: utf-8 -*-
import socket
import sys
import libvirt
if (__name__ == "__main__"):
   conn = libvirt.open("qemu+ssh://xxx/system")
   print "Trying to find node on xxx"
   domains = conn.listDomainsID()
   for domainID in domains:
       domConnect = conn.lookupByID(domainID)
       if domConnect.name() == 'xxx-node':
           print "Found shared node on xxx with ID " + str(domainID)
           domServ = domConnect

Bridged Networking

To use physical Ethernet from your virtual machines, you have to create a bridge between your physical Ethernet device (here eth0) and the virtual Ethernet device the VM is using.

Host configuration

libvirt creates the bridge virbr0 for NAT networking, so use another name such as br0 or virbr1. You have to create a new Netcfg Profile to configure the bridge, for example (with DHCP configuration):

DESCRIPTION="KVM Bridge connection"
## sets forward delay time
## sets max age of hello message
Tip: It is recommended that you enable Spanning Tree Protocol (STP) on the virtual bridge (e.g. br0) that you create to avoid any potential bridging loops. You can automatically enable STP by appending POST_UP="brctl stp $INTERFACE on" to the netcfg profile.

Guest configuration

Now we have to activate the bridge interface in our VMs. If have a recent Linux machine, you can use this code in the .xml file:

 <interface type='bridge'>
   <source bridge='br0'/>
   <mac address='24:42:53:21:52:49'/>
   <model type='virtio' />

This code activates a virtio device on the machine so, in Windows you will have to install an additional driver (you can find it here Windows KVM VirtIO drivers) or remove the line <model type='virtio' />:

 <interface type='bridge'>
   <source bridge='br0'/>
   <mac address='24:42:53:21:52:49'/>