Translation(s): ?Indonesian

(!) ?/Discussion


Xen Overview

Modern computers are sufficiently powerful to use virtualization to present the illusion of many smaller virtual machines (VMs), each running a separate operating system instance. Successful partitioning of a machine to support the concurrent execution of multiple operating systems poses several challenges. Firstly, virtual machines must be isolated from one another: it is not acceptable for the execution of one to adversely affect the performance of another. This is particularly true when virtual machines are owned by mutually untrusting users. Secondly, it is necessary to support a variety of different operating systems to accommodate the heterogeneity of popular applications. Thirdly, the performance overhead introduced by virtualization should be small.

Xen is a virtual machine monitor for x86 that supports execution of multiple guest operating systems with unprecedented levels of performance and resource isolation. Xen is Open Source software, released under the terms of the GNU General Public License. We have a fully functional ports of Linux 2.6 running over Xen, and regularly use it for running demanding applications like MySQL, Apache and PostgreSQL. Any Linux distribution (RedHat, SuSE, Debian, Mandrake) should run unmodified over the ported OS.

In addition to Linux, members of Xen's user community have contributed or are working on ports to other operating systems such as NetBSD (Christian Limpach), FreeBSD (Kip Macy) and Plan 9 (Ron Minnich).

Different types of virtualization offered by Xen

There are two different types of virtualization offered by Xen:

Para-virtualization

Para-virtualization, also known as software virtualization or PV, is the type one uses when the kernel of the dom0 is freely distributed as source code. That way, modifications and patches to the kernel can be made, to allow the Linux kernel to be aware of the multiple domains

Hardware Virtual Machine

The other technique Xen uses is the Hardware Virtual Machine. It is also known as HMV. HMV is useful and required when there is no source access to the kernel of the Operating System. This way, one can run a host operating system virtualized, which no modifications. CPUs that support this hardware virtualization, provide special registers for domU to run on, in order to minimize any loss in performance.

It is also possible to run some other, not ported, operating systems if your CPU has hardware virtualization support (VT or Pacifica).

Compatibility

Installation on lenny

The short of the story is the following:

  1. Install Lenny. Finish the installation as you would normally do
  2. apt-get install install xen
  3. apt-get install xen-utils
  4. apt-get install xen-tools
  5. install a xen dom0-capable kernel
  6. reboot into xen
  7. create your domains
  8. manage your domains

Dom0 (host)

After you are done installing the base OS, xen-utils and xen-tools, you will need to install a xen-capable dom0 kernel. The kernel is 2.6.26, the -xen variant contains patches from SuSE for dom0 support.

The xen-linux-system packages of interest are (Install the correct one for your architecture):

Serial console access

To get output from grub, XEN, the kernel and getty (login prompt) via both vga and serial console to work, here's an example of the right settings when using Lenny kernels and Xen 3.2:

In /boot/grub/menu.lst:

serial --unit=0 --speed=9600 --word=8 --parity=no --stop=1
terminal --timeout=5 serial console
[...]
title           Xen 3.2-1-amd64 / Debian GNU/Linux, kernel 2.6.26-2-xen-amd64
root            (hd0,0)
kernel          /boot/xen-3.2-1-amd64.gz com1=9600,8n1 console=com1,vga
module          /boot/vmlinuz-2.6.26-2-xen-amd64 root=/dev/md0 ro console=tty0 console=hvc0
module          /boot/initrd.img-2.6.26-2-xen-amd64

In contrast to the Etch configuration, there's no ttyS0 in the vmlinuz line!

In /etc/inittab you need at least these lines:

1:2345:respawn:/sbin/getty 38400 hvc0
2:23:respawn:/sbin/getty 38400 tty1
# NO getty on ttyS0!

The tty1 will show up at the vga output, and the hvc0 will show up at the serial console.

DomU (guest)

The Lenny Debian Installer fully supports installation of 32 bit guests under Xen using the netboot/xen variant. Images are available on any Debian mirror in the installer directory and contain a kernel, installer ramdisk and an example Xen configuration file. To install, fetch the xm-debian.cfg configuration file, edit to suit your tastes, and start the guest with the install=true option plus an optional (but strongly recommended) install-mirror=ftp://ftp.XX.debian.org/debian.

xm create -c xm-debian.cfg install=true install-mirror=ftp://ftp.XX.debian.org/debian

Newer images are also available from the daily builds. After grabbing the xm-debian.cfg configuration file and editing it to suit your tastes, start the guest with an additional install-installer=http://people.debian.org/~joeyh/d-i/images/daily/ to manually direct it to the daily builds:

WARNING, if you do not change the hard disks option on xm-debian.cfg this WILL overwrite your dom0 instead of installing to your domU. YOUR MACHINE WILL BE DESTROYED.

xm create -c xm-debian.cfg install=true \
  install-mirror=ftp://ftp.XX.debian.org/debian \
  install-installer=http://people.debian.org/~joeyh/d-i/images/daily/

See the comments in the configuration file for additional installation options.

Another way of creating a lenny domu is the following:

xen-create-image --hostname=vanila --size=8Gb --dist=lenny --memory=512M --ide --dhcp

Please note that the --dir option may be required, and it specified the directory where it will store your disk images. If you wish to specify a fixed ip address, use the --ip xxx.xxx.xxx.xxx instead of --dhcp option.

Once the guest is installed simply boot it using:

xm create -c xm-debian.cfg

Lenny only includes 32 bit (PAE) kernel support which means there is no installer support for 64 bit guests. You can continue to use the Etch kernels or obtain a newer upstream kernel which supports 64 bit operation (2.6.27+).

In addition to installing via Debian Installer xen-tools can also create a Lenny domU as described in the Etch section above.

The default Lenny kernel is the newer paravirt_ops version (2.6.26), which does not function as a dom0 (except for the -xen variants, which have dom0 support but also some issues running as domU (please clarify?). It will also not support PCI passthrough in a domU. For PCI passthrough, you have to run the 2.6.18 etch kernel as both dom0 and domU.

In Lenny the distinction between the Xen and non-Xen flavours of the kernel (with respect to domU support) is no longer present. The Debian Installer will install the -686-bigmem flavour.

Additional note for domU on lenny using xen-tools

xen-tools don't use hvc0 as the console device in /etc/inittab and don't install udev (leading to /dev/pts missing in domU).

This makes logging in via xm console and via ssh impossible, because getty doesn't have a proper console to attach to and ssh can't attach to a pseudo terminal.

To fix this,

1. add to /etc/xen-tools/xen-tools.conf:

serial_device = hvc0

2. and make domU with:

xen-create-image --hostname HOSTNAME (more options...) --role udev

Installation on etch

Upstream documentation can be found in the xen-docs-3.0 package (in /usr/share/doc/xen-docs-3.0/user.pdf.gz). It's also available online.

Dom0 (host)

The xen-linux-system packages of interest are (Install the correct one for your architecture):

If you need to apply some modifications to the kernel with the xen patch, then one way to do it is described ?DebianKernelCustomCompilation.

Serial console access

To get output from grub, XEN, the kernel and getty (login prompt) via both vga and serial console to work, here's an example of the right settings when using etch kernels and Xen 3.0.3:

In /boot/grub/menu.lst:

serial --unit=0 --speed=9600 --word=8 --parity=no --stop=1
terminal --timeout=5 serial console
[...]
title           Xen 3.0.3-1-i386-pae / Debian GNU/Linux, kernel 2.6.18-6-xen-686
root            (hd0,0)
kernel          /boot/xen-3.0.3-1-i386-pae.gz com1=9600,8n1 console=com1,vga
module          /boot/vmlinuz-2.6.18-6-xen-686 root=/dev/md0 ro console=tty0 console=ttyS0,9600n8
module          /boot/initrd.img-2.6.18-6-xen-686

In /etc/inittab you need at least these lines:

1:2345:respawn:/sbin/getty 38400 tty1
T0:23:respawn:/sbin/getty -L ttyS0 9600 vt100

DomU (guests)

The easiest way to create a domU is to use xen-tools (and, if this doesn't do what you need, Steve Kemp is keen and fast in implementing useful suggestions).

If you do not wish to use xen-tools, you could use this alternative guide, to setup the system using debootstrap.

Xen boots domUs using kernels stored on dom0, so you only need to install the corresponding linux-modules package in the domU. Alternatively, you can use PyGrub to boot kernels on the domU filesystem.

On i386, make sure you install libc6-xen.

If you install lenny domU on etch dom0, make sure you read this entry on XenFaq when you see messages on the console like 4gb seg fixup, process klogd (pid 2075), cs:ip 73:b7e25870. After applying the echo 'hwcap 0 nosegneg' > /etc/ld.so.conf.d/libc6-xen.conf && ldconfig, in the dom0 system, reboot, or, if you don't like rebooting (which requires you to stop domU's), restart all processes mentioned in the log messages (e.g. /etc/init.d/ssh restart, init q, etc..)

Xen on Testing/Squeeze and on Unstable/Sid

Steps

  1. apt-get install linux-image-2.6.XX-X-xen-amd64
  2. apt-get install xen-tools
  3. apt-get install xen-utils
  4. reboot
  5. xen-create-image --hostname my-xen1 --dhcp --dir /home/s1-xen1/

Geeky Notes

The new 2.6.32 kernel images have paravirt_ops-based Xen dom0 and domU support, please use them:

For domU, you can use either the same packages that support dom0 (those that include '-xen-' in their names), or the linux-image-2.6-686-bigmem and linux-image-2.6-amd64 packages.

For those who want to test the 2.6.32 kernel domU on an earlier dom0, you have to make sure that the xen-blkfront domU driver will be able to find the root and other disk partitions, which is no longer the case if you still use the deprecated hda* or sda* device names in domU .cfg files. Switch to xvda* devices, which also work with 2.6.18 and 2.6.26 dom0 kernels.

There are also the backward-looking options:

Bugs You May Encounter

Debian Bug #584152 Error during xen-create-image: mkfs.ext3: /lib/libblkid.so.1: version `BLKID_2.17' not found (required by mkfs.ext2). Solve this by downgrading the mkfs tool. Xen Bug #1620 Error starting xend.

Using Debian-Installer

The page ?DebianInstaller/Xen contains instructions on how to install Xen Dom0 and Etch DomU with DebianInstaller.See above for details of installing Lenny using Debian Installer.

Package maintenance

Debian's Xen packages are maintained by the pkg-xen project. (developpers' mailing list)

The Debian Developer's Package Overview page lists source packages that are maintained by the team.

Common Errors

dom0 automatic reboots

Error "Device ... (vif) could not be connected"

You need to configure some basic networking between dom0 and domU. Edit /etc/xen/xend-config.sxp

#(network-script network-dummy)
(network-script network-bridge)

for a basic bridge networking, and restart xend.

error: CDROM boot failure

You get the error :

That's because Xen can't boot from a cdrom iso image at the moment. i.e you can't have tap:aio:/path/to/mycd.iso,hdc:cdrom,r or file:/path/to/mycd.iso,hdc:cdrom,r.

Workaround: use losetup to create a loopback device for the cdrom ISO image, then use it in Xen configuration file. for example :

Now you can use /dev/loop9 in xen configuration file (/etc/xen/foobar.cfg) :

then boot/install the guest OS.

note: yo should switch back to the tap:aio:/path/to/mycd.iso,hdc:cdrom,r syntax after installation, since loop back have to be recreated after you reboot the host system.

4gb seg fixup errors

Solution:

echo 'hwcap 0 nosegneg' > /etc/ld.so.conf.d/libc6-xen.conf && ldconfig

Read this XenFaq entry for more info.

No login prompt when using `xm console`

Using a lenny domU, make sure you have hvc0 listed in inittab, like 1:2345:respawn:/sbin/getty 38400 hvc0. There happened to be a lot of changes of default console unit used by Xen (tty1, xvc0, hvc0 etc) but for a Lenny domU (version > 2.6.26-9) it's hvc0.

'clocksource/0: Time went backwards'

If a domU crashes or freezes while uttering the famous lasts words 'clocksource/0: Time went backwards', your domU is likely using the xen clocksource instead of its own clock ticks. In practice, this seems to be the cause of infrequent lockups under load (and/or problems with suspending).

see http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1098

workaround #1

A workaround is to decouple the clock in the domU from the dom0:

In your dom0 and domU /etc/sysctl.conf add the line: xen.independent_wallclock=1. On the dom0, edit the configuration file of the domU (e.g. /etc/xen/foobar.cfg and add (or expand) the extra-line: extra="clocksource=jiffies".

These settings can be activated without rebooting the domU. After editing the configuration files, issue sysctl -p and echo "jiffies"> /sys/devices/system/clocksource/clocksource0/current_clocksource on the domU prompt.

Because the clock won't be relying on the dom0 clock anymore, you probably need to use ntp on the domU to synchronize it properly to the world.

workaround #2

Another possibility ist to use the behaviour of the previous xen-kernel settings: clocksource=jiffies and independent_wallclock=0

Setting clocksource=jiffies for the dom0 and each domU as kernel parameter has eliminated the "Time went backwards" for me (14 dom0s and 27 domUs running stable for two weeks). You can check the values with

cat /sys/devices/system/clocksource/clocksource0/current_clocksource

and

cat /proc/sys/xen/independent_wallclock

With these settings, ntp ist only needed in the dom0. If you change the time in a domU while ntp is running on the according dom0, time will be corrected within a few minutes in the domU. Hint: I didn't manage to influence the time of the domU with setting the time in the dom0 with date or hwclock, nevertheless ntp seems to do this (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=534978#29).

"Error: Bootloader isn't executable"

The above, rather cryptic, error (when starting a domU using xen-utils/xm create) is due to xen-utils not being able to find PyGrub. Modify your xm-debian.cfg config file to use the absolute directory (ie. bootloader="/usr/lib/xen-3.2-1/bin/pygrub" instead of bootloader="pygrub") and your domU should boot up fine.

"ERROR (XendCheckpoint:144) Save failed on domain mydomu32 (X)."

xm save/migration of a 32-Bit domU on a 64-Bit dom0 fails. It seems this is not supported with linux-image-2.6.26-2-xen-amd64 (http://readlist.com/lists/lists.xensource.com/xen-users/4/24225.html). One workaround is to use a 64-Bit Hypervisor with a 32-Bit dom0 (http://lists.xensource.com/archives/html/xen-users/2008-12/msg00404.html). See also 526695

"network routing for hvm guests:"

ERROR in /var/log/xen/qemu-dm-[.*].log:

bridge xenbr0 does not exist!

/etc/xen/scripts/qemu-ifup: could not launch network script

When using routing instead of bridging there seems to be problems for hvm guests. Here a very bad hack for it: prerequsites:

in "/etc/xen/xend-config.sxp"

in your domU config file

than:

In "/etc/xen/scripts/qemu-ifup" disable with a #

insert

after starting you domU

pretty bad but works...

PV drivers on HVM guest

It may be possible to build the PV drivers for use on HVM guests. These drivers are called unmodified_drivers and are part of the xen-unstable.hg repository. You can fetch the repository using mercurial thus:

The drivers reside under xen-unstable.hg/unmodified_drivers/linux-2.6. The README in this directory gives compilation instructions.

NUMA with xen 3.4

In order to activate NUMA awareness in the hypervisor on multi-socket AMD and Intel hosts, use the following:

acpi=on numa=on

by default, NUMA is off.

Resources


CategoryNetwork