Blue Light
Space shortcuts
Child pages
  • Xen
Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 21 Next »

Introduction

Xen is Blue Light's standard virtualisation solution, decided in BLUE-1192

Introduction to Xen: http://wiki.xenproject.org/wiki/Xen_Beginners_Guide#What_is_this_Xen_Project_software_all_about.3F

Installation (host/dom0)

TODO: better/different to install xen-linux-system instead of xen-tools?

aptitude install xen-linux-system

Adjust grub so the default boot item is Xen (ref: https://xen-orchestra.com/cant-find-hypervisor-information-in-sysfs/)

dpkg-divert --divert /etc/grub.d/08_linux_xen --rename /etc/grub.d/20_linux_xen
update-grub
shutdown -r now

Test: does the  xen list  command show that the current system is Domain-0?

Storage

The recommended way to provide storage for a DomU is by an LVM on Dom0 which the DomU uses as a virtual HDD.

Reference: http://wiki.xenproject.org/wiki/Storage_options

Networking

Choice (TODO: can be mixed?):

  • Bridged  Each DomU* OS is connected to the LAN
  • Routed  Dom0 is a router between the LAN and the DomU* network interfaces
  • NATted  Dom0 is a NATting router between the LAN and the DomU* network interfaces

Interface names:

  • peth*  The physical ethernet interfaces in a Xen bridge, connected to the LAN
  • vif*  The virtual interfaces in a Xen bridge, connected to the DomUs
  • veth*  TBC
  • xenbr*  A Xen bridge

Bridged

Dom0's eth0 gets the bridge's address, broadcast, netmask and  gateway settings (assuming the bridge config includes bridge_ports eth0).

Diagram from http://wiki.xenproject.org/wiki/Xen_Networking#ASCII_Art_Examples_of_Xen_Networking_Topologies:

      LAN0                                                  LAN1
       |                                                     |
 +-----+-----------------------------------------------------+-----+
 |     |                                                     |     |
 | +---+-------------------------+ +-------------------------+---+ |
 | |   |                         | |                         |   | |
 | | eth0                        | |                        eth1 | |
 | |                             | |                             | |
 | | xenbr0       vif1.0  vif2.0 | |  vif1.1  vif2.1      xenbr1 | |
 | |                |       \    | |    /       |                | |
 | +---^------------+---------\--+ +--/---------+------------^---+ |
 |     |            |           \   /           |            |     |
 |     |     +------+-------------X-------------+------+     |     |
 |     |     |      |           /   \           |      |     |     |
 |     |     | +----+---------/--+ +--\---------+----+ |     |     |
 |     |     | |    |       /    | |    \       |    | |     |     |
 |     |     | |  eth0    eth1   | |   eth0   eth1   | |     |     |
 |     |     | |    |       |    | |    |       |    | |     |     |
 |   +-+-+   | |  +-+-+   +-+-+  | |  +-+-+   +-+-+  | |   +-+-+   |
 |   |   |   | |  |   |   |   |  | |  |   |   |   |  | |   |   |   |
 |  www ssh  | | www ssh ftp pop | | www ssh ftp pop | |  ftp pop  |
 |           | |                 | |                 | |           |
 |  Domain0  | |     Domain1     | |     Domain2     | |  Domain0  |
 +-----------+ +-----------------+ +-----------------+ +-----------+

Useful commands:

  • brctl show

Routed

When routing is used, need:

net.ipv4.ip_forward = 1
net.ipv4.conf.eth0.proxy_arp = 1

NATted

When NAT is used, need:

iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

A cunning plan?

Maybe we can do what we want (as on blav; some VMs with public IPs, most with NATted non-routable IPs) without a bridge; configure the physical interface with all the public IPs and use IPTABLES to forward traffic sent to notionally the VM's public IPs to the VM's non-routable IPs.  For each VM adressable by public IP:

iptables -t nat -A PREROUTING -i eth0 -d PUBLICIP -j DNAT --to-destination INTERNALIP

For NATting:

iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

References (network)

Toolstacks (a.k.a toolkits and toolboxes)

A toolstack is a set of user-space tools used for adminstering Xen.

Several toolstacks are available for the interactive or scripted administration of Xen resources.

Primary reference: http://wiki.xen.org/wiki/Choice_of_Toolstacks

Apparently, once a toolstack has been chosen, it is not easy to change.  From Debian 7 Wheezy's /etc/default/xen (where the chosen toolstack is configured):

# Attention: You need to reboot after changing this!

libvirt and virsh

libvirt is a library for managing KVM, OpenVZ, VMware, VirtualBox, Xen and others.  For Xen, it is the least feature-complete of all the toolstacks.

virsh is a user/script shell to interface with libvirt.  References: http://libvirt.org/virshcmdref.html

xapi and xe

The most feature-complete of all the toolstacks.

xend and xm

Deprecated since Xen 4.1 but the default in Xen 4.1 as packaged for Debian.  Will be removed from Xen 4.2.

xend comprises:

  • xm command

  • xend-config.sxp config file format

  • xmdomain.cfg config file format

xl

Strategic since Xen 4.1.

Designed to be command line compatible with xend.  TODO: what does this mean?  man -k xend lists the config file format xend-config.sxp.  It may mean command line compatible with xm; many of the xl commands take similar arguments to xm.

Unlike xm, xl does not do dom0 network configuration; standard OS tools must be used.

In case there is no local xl man page: http://xenbits.xen.org/docs/unstable/man/xl.1.html and (Xen 4.2) http://manpages.ubuntu.com/manpages/raring/en/man1/xl.1.html

Toolstack library

Libxenlight (libxl)

The strategic Xen toolstack library, designed to hide xenstore, libxenctrl, and libxenguest from higher levels.

Toolstack configuration

Debian 7 Wheezy

The Xen toolstack on Debian 7 Wheezy is configurable.

On blav2, we had explicitly configured using xl:

root@SON-OF-blav.bluelightav:~# diff /etc/default/xen{.org,}
7c7 < TOOLSTACK=
--- > TOOLSTACK=xl

When root runs command xl:

  1. The shell runs /usr/sbin/xl which is a symlink to /usr/lib/xen-common/bin/xen-toolstack-wrapper
  2. /usr/lib/xen-common/bin/xen-toolstack-wrapper:
    1. Sources /usr/lib/xen-common/bin/xen-toolstack to set TOOLSTACK to /usr/lib/xen-4.1/bin/xl
    2. Sources /usr/lib/xen-common/bin/xen-dir to set dir to /usr/lib/xen-4.1
    3. Sources /usr/lib/xen-common/bin/xen-version to set VERSION to 4.1
      1. /usr/lib/xen-common/bin/xen-version gets Xen version info from /sys/hypervisor/version/*
    4. Sets PATH to /usr/lib/xen-4.1/bin
    5. Sources /etc/default/xen to set TOOLSTACK to xl (would fall back to hardcoded default otherwise)
    6. Sets TOOLSTACK to /usr/lib/xen-4.1/bin/xl
    7. execs /usr/lib/xen-4.1/bin/xl

When root runs command xm (does not work):

  1. The shell runs /usr/sbin/xm which is a symlink to /usr/lib/xen-common/bin/xen-toolstack-wrapper
  2. /usr/lib/xen-common/bin/xen-toolstack-wrapper:
    1. Sources /usr/lib/xen-common/bin/xen-toolstack to set TOOLSTACK to /usr/lib/xen-4.1/bin/xl
    2. Sources /usr/lib/xen-common/bin/xen-dir to set dir to /usr/lib/xen-4.1
    3. Sources /usr/lib/xen-common/bin/xen-version to set VERSION to 4.1
      1. /usr/lib/xen-common/bin/xen-version gets Xen version info from /sys/hypervisor/version/*
    4. Sets PATH to /usr/lib/xen-4.1/bin
    5. Sources /etc/default/xen to set TOOLSTACK to xl (would fall back to hardcoded default otherwise)
    6. Sets TOOLSTACK to /usr/lib/xen-4.1/bin/xl
    7. Generates message "ERROR: A different toolstack (xl) have been selected!"

Maintaining Debian DomUs

This is listed before "Creating Debian DomUs" because a newly created Debian should be updated during installation.

Installing apt-cacher on Dom0 will save bandwidth and speed maintenance (and require significant time cleaning broken cache files).

Creating Debian DomUs

debootstrap

Installs into a chroot.  Can use only one repository (which may be a local cache).

References:

Debian installer

"Debian installer" is the standard Debian installation system; it supports installation of standard Debian as a Xen PV.

Pre-seeding

If we use either of the Debian installer methods, pre-seeding would be advantageous both for creating installations that differ only as intended and – more importantly – to accelerate re-installation during disaster recovery.  This might be very useful for all Debian and ubuntu installations, not just DomUs.

Reference: http://www.debian.org/releases/stable/amd64/apb.html.en.

Actually FAI (http://fai-project.org/ and BLUE-1325) is the way to go but it will take too long to familiarise with it.

Even pre-seeding is non-trivial so decided not to delay the first installation by using it.

References:

Netboot

http://www.debian.org/releases/stable/amd64/ch04s05.html.en

.iso image

Installing from a .iso image is the method we are most familiar with so a good choice, at least for the first DomUs.

Installation method decision

24 July 2014

Decided to install from a .iso rather than any of the other choices because:

  • We are already familiar with it
  • Having no integration into the Xen architecture, we will learn more about Xen during the domU process.

xen-create-image (from xen-tools)

xen-create-image is part of the xen-tools package.

When creating Debian systems it uses debbootsrap (described above).

Troubleshooting

TODO: integrate the next para.

A quick note, there are actually 3 modes, not two when it comes to the drivers in use:

  • HVM: unmodified kernel and drivers using software emulated devices
  • PV-HVM: unmodified kernel with paravirtualized (Xen specific) disk and network drivers
  • PV: modified kernel and drivers

For a Xen guest/DomU you can do a very basic uname and lsmod with a grep to list the modules in use:

uname -a lsmod | grep xen

If uname -a lists a kernel with the string "xen" in it, then you have a modified kernel and it's likely a PV guest, and you will see output from the lsmod command to confirm it. If you have output from the grep on lsmod but no sign of a modified kernel then you are PV-HVM. Without any sign of either, it's a straight HVM.

Note: Generally you can do more with VMs that have the PV tools installed, so that can be quite an obvious pointer, however you can fake the presence of the PV tools to allow suspend/resume etc. so you cannot rely on that in general.

References

Files and directories

File or directoryUsageNotes
/etc/default/xen  
/etc/default/xend  
/etc/default/xendomains  
​/etc/xen/​Configuration​
/etc/xen/scripts/? 
/etc/xen/scripts/vif-nat  

Glossary

  • Domain  a Xen VM.
  • dom0  The Xen domain/VM used a) to control the hypervisor b) as the primary interface to the hardware.
  • domu*  Unprivileged Xen domains/VMs
  • Enlightened guest  An OS designed/modified for use in a domain/VM.
  • Full virtualisation  A domain/VM providing full emulation of physical hardware.  The OS can be the same as runs on real hardware.
  • Guest domain  Same as domu*.
  • Hardware Virtual Machine (HVM)  Same as full virtualisation.
  • Hypervisor a software system that allows the execution of multiple virtual guest operating systems simultaneously on a single physical machine. Xen's hypervisor is Type 1 or “bare-metal”, meaning it runs on the physical machine as opposed to within an operating system.
  • HVM  Hardware Virtual Machine.
  • Para-virtualisation  uses modified guest operating systems a.k.a. enlightened guests. They don’t require virtual hardware devices, instead they make special calls to the hypervisor that allow them to access CPUs, storage and network resources.  Better performance that full virtualisation.
  • PV  para-virtualisation.
  • Virtual Machine Monitor (VMM)  Same as hypervisor.
  • Xen Store  A database of information shared by the hypervisor, the kernels, the drivers and the xen daemon.  The memory for Xen Store is provided by Dom0.  In Dom0 (and in enlightened guests?) it can be accessed via /proc/xen/xenbus/

Technical

TODO; move Xen notes from the Online.net server page to here.

Architecture

Very little information found on the 'net about Xen architecture.  This diagram is from http://libvirt.org/architecture.html

TODO: mine http://wiki.xen.org/wiki/XenIntro for more information.

Primary reference (no Xen 4.x equivalent found): http://wiki.xen.org/wiki/XenIntro

Debian packages

libxen-4.1

Includes /usr/lib/libxenctrl-4.1.so and /usr/lib/libxenguest-4.1.so

libxenstore3.0

Includes /usr/lib/libxenstore.so.3.0.0 and /usr/lib/libxenstore.so.3.0

xen-hypervisor-4.1-amd64

Includes /boot/xen-4.1-amd64.gz

xen-linux-system-3.2.0-4-amd64

Documentation only.  TODO: presumably its dependencies are crucial; list them?

xen-linux-system-amd64

Documentation only.  TODO: presumably its dependencies are crucial; list them?

xen-system-amd64

Documentation only.  TODO: presumably its dependencies are crucial; list them?

xen-tools

Tools that work with the xm and xl toolstacks to administer Xen config files and images.

Provides (in /usr/bin):

  • xt-customize-image
  • xt-create-xen-config
  • xen-list-images
  • xen-delete-image
  • xt-install-image
  • xen-create-nfs
  • xen-create-image
  • xt-guess-suite-and-mirror
  • xen-update-image

Plus, presumably in suppport:

  • /etc/xen-tools/*
  • /usr/lib/xen-tools/<OS name>.d/*

xen-create-image largely automates creating domUs.  Config file: /etc/xen-tools/xen-tools.conf.  The --lvm option creates a new logical volume which is available to the domU as a hard disk drive

Logs in the /var/log/xen-tools/ directory.

References

xen-utils-4.1

Tools to control the hypervisor from dom0.

Provides (in /usr/lib/xen-4.1/bin/):

  • xentrace_setsize
  • xenpm
  • xen-python-path
  • qemu-dm
  • xsview
  • xen-hvmcrash
  • xentrace_setmask
  • xenstored
  • xentrace_format
  • xenbaked
  • readnotes
  • xenconsoled
  • xc_restore
  • gtraceview
  • xen-bugtool
  • xen-hptool
  • xentop
  • xend
  • xc_save
  • gtracestat
  • xenctx
  • xenwatchdogd
  • xenpaging
  • pygrub
  • xencons
  • xen-detect
  • lsevtchn
  • tap-ctl
  • xl
  • xentrace
  • xm
  • xen-hvmctx
  • xenconsole
  • xenstore-control
  • xenmon
  • xenlockprof
  • xen-tmem-list-parse
  • xenperf

Plus, presumably in suppport:

  • /usr/lib/xen-4.1/lib/
  • /usr/lib/xen-4.1/boot/hvmloader

xen-utils-common

Provides:

  • /etc/default/xend
  • /etc/default/xendomains
  • /etc/init.d/xen
  • /etc/init.d/xendomains
  • /etc/xen/scripts/*
  • /etc/xen/xend-config.sxp
  • /etc/xen/xend-pci-permissive.sxp
  • /etc/xen/xend-pci-quirks.sxp
  • /etc/xen/xl.conf
  • /lib/udev/rules.d/xen-backend.rules
  • /lib/udev/rules.d/xend.rules
  • /usr/lib/xen-common/bin/*
  • /usr/sbin/cpuperf-perfcntr
  • /usr/sbin/cpuperf-xen
  • /usr/sbin/xe
  • /usr/sbin/xenperf
  • /usr/sbin/xenpm
  • /usr/sbin/xentop
  • /usr/sbin/xentrace
  • /usr/sbin/xentrace_format
  • /usr/sbin/xentrace_setmask
  • /usr/sbin/xentrace_setsize
  • /usr/sbin/xl
  • /usr/sbin/xm
  • /usr/share/xen-utils-common/default.xen

xenstore-utils

Provides (all in /usr/sbin):

  • xenstore-ls
  • xenstore-write
  • xenstore-exists
  • xenstore-rm
  • xenstore-chmod
  • xenstore-watch
  • xenstore-list
  • xenstore-read

References

General references are listed here.  Specific references are listed in the sections they apply to above.

Introductory

Entry points

These are entry points to references.

Setup logs

 

  • No labels