XenServer Documentation

This chapter provides an overview of how VMs are created and lists virtual memory and virtual disk size minimums, describes the differences in virtual device support for the members of the XenServer product family. This chapter also discusses physical to virtual conversion (P2V), cloning templates, and importing previously-exported VMs.

Overview

VMs are created from templates. A template is a "gold image" that contains all the various configuration settings to instantiate a specific VM. XenServer ships with a base set of templates, which range from generic "raw" VMs that can boot an OS vendor installation CD (Windows) or run an installation from a network repository (Red Hat Enterprise Linux, SUSE Linux Enterprise 10) to complete pre-configured OS instances (Debian Etch and Sarge).

There are three basic methods by which VMs are created using templates:

using a complete pre-configured template (Debian Sarge and Etch Linux)
Installing from a CD or an ISO image onto the appropriate template (Windows 2000 SP4/Windows 2003 Server/Windows XP SP2/Windows Vista, RHEL 5.0, CentOS 5.0)
Installing from vendor media on a network installation server directly onto a template (Red Hat Enterprise Linux 4.x and 5.0, and SUSE Linux Enterprise Server 10 SP1)

Creating VMs by installing Windows operating systems onto the appropriate templates is described in Installing Windows VMs.

Creating VMs by installing Linux operating systems onto the appropriate templates is described in Installing Linux VMs.

Additionally, VMs can be created by

performing a physical to virtual (P2V) conversion on an existing physical server (Red Hat Enterprise Linux 3.6, 3.8, 4.1-4.4, and SUSE Linux Enterprise Server 9 SP2 and SP3)
importing an existing, exported VM
converting an existing VM to a template

These methods are describe in this chapter.

Virtual memory and disk size limits

In general, when installing VMs, be sure to follow the memory and disk space guidelines of the operating system and any relevant applications that you want to run when allocating resources such as memory and disk space.

Note that individual versions of the operating systems may also impose their own maximum limits on the amount of memory supported (for example, for licensing reasons). Also, since XenServer Express Edition only supports a maximum of 4GB of physical RAM for the XenServer Host, 4GB is the upper limit for a VM on this platform.

Operating System	Minimum RAM	Maximum RAM	Disk space
Windows Vista 32-bit	512MB minimum supported, 768MB or more recommended	32GB	16GB
Windows 2003	128MB minimum supported; 256MB or more recommended	32GB	2GB
Windows XP SP2	128MB minimum supported; 256MB or more recommended	32GB	1.5GB
Windows 2000 SP4	128MB minimum supported; 256MB or more recommended	32GB	2GB
CentOS 4	256MB	16GB	800MB
CentOS 5	512MB	16GB	800MB
Red Hat Enterprise Linux 3.6	64MB	32GB	1.5GB
Red Hat Enterprise Linux 4.1, 4.4	256MB	32GB	800MB
Red Hat Enterprise Linux 4.5, 4.6	256MB	16GB	800MB
Red Hat Enterprise Linux 5.0, 5.1	512MB	16GB	800MB
SUSE Linux Enterprise Server 9 SP2	256MB	32GB	1GB
SUSE Linux Enterprise Server 10 SP1	512MB	32GB	1.5GB
Debian Sarge, Etch	128MB	32GB	4GB

XenServer product family virtual device support

The current version of the XenServer product family has the following general limitations on virtual devices for VMs. Note that specific guest operating systems may have lower limits for certain features. These limitations are noted in the individual guest installation section.

Virtual device	Linux VMs	Windows VMs
Number of virtual CPUs	32^[a]	8
Number of virtual disks	8 (including virtual CD-ROM)	8 (including virtual CD-ROM)
Number of virtual CD-ROM drives	1	1
Number of virtual NICs	7^[b]	7
Hotplugging virtual disks	add/remove	add/remove
Hotplugging virtual NICs	add/remove	add/remove
^[a]A maximum of 8 vCPUs are supported via XenCenter. ^[b]except for SLES 10 SP1 and RHEL 3.x, 4.x, and 5.x, which support 3

Express Edition, Standard Edition, and Enterprise Edition also differ in the following ways that are relevant for creating VMs:

	Enterprise Edition	Standard Edition	Express Edition
Amount of physical RAM on XenServer Host	up to 128GB	up to 128GB	up to 4GB
Number of concurrent VMs	50	50	4
Support for VLANs	yes	yes	no
Support for shared storage	yes	no	no
Support for server pools	yes	no	no
Support for additional QoS control	yes	no	no

If you attempt to create a fifth VM with a Express Edition license, for example, an error message will appear suggesting that you can upgrade your license. Licenses are applied per-host, and so for a resource pool setup you must apply a Enterprise Edition license across all hosts before joining them into a pool.

If you downgrade a license on a host, it does not take any immediate action on running domains, but ensures that the restrictions are enforced from that point onwards (e.g. if you have 5 VMs running and downgrade to Express Edition, they will continue to run but you cannot start any more). License downgrade is disallowed in the case of a host that is actively participating in a pool, so it must be ejected and then downgraded.

License expiry on a host also does not take any immediate action on running domains, but prevents new domains being started. XenCenter will also regularly warn you if your hosts are approaching license expiry thresholds, in advance of it.

Physical to Virtual Conversion (P2V)

Physical to Virtual Conversion (P2V) is the process by which an existing operating system on a physical server — its filesystem, configuration, etc. — is cast into a virtualized instance of the same operating system and filesystem, transferred, instantiated, and started as a VM on the XenServer Host.

For existing physical instances of Windows servers, a third-party tool is required. Windows P2V software and documentation is available for download from the Citrix website at http://www.citrixxenserver.com/partners/Pages/PartnerOffers.aspx.

For existing physical instances of Linux servers, this is accomplished by booting from the XenServer installation CD and choosing the P2V option. The filesystem is copied across the network onto a XenServer Host, where it appears as a normal VM. We recommend that you perform P2V operations during off-peak hours since the process involves transferring a large amount of data, which could impact other Virtual Machines running on the XenServer Host.

The P2V tool requires a 64-bit capable CPU by default. If you have an existing Linux instance on an older machine that you want to transfer via P2V, you can boot the CD via the p2v-legacy option at the initial prompt. This does require at least a PAE-enabled machine, so for very old machines you can physically move the hard drive to a PAE-enabled machine and perform the operation from there.

To P2V an existing Linux server directly to a XenServer Host

Reboot the physical server that you want to convert and boot from the XenServer installation CD. If the boot fails, start again and use the p2v-legacy option.
After the initial boot messages, you will see the "Welcome to XenServer" screen. (In this and the screens that follow, use Tab or Alt+Tab to move between elements, Space to select, and F12 to move to the next screen.)Select OK to proceed.
The installer does some hardware detection and initialization, then presents a screen with four choices.Select Convert an existing OS on this machine to a VM (P2V) and choose OK to proceed.
A "Welcome to XenServer P2V" screen with a descriptive message is displayed next. Click OK to proceed and follow the on-screen prompts.

When the P2V process is complete and the new VM is created, you will need to create and attach a VIF for it to have external network connectivity. Similarly, extra disks may also be added to take advantage of additional storage capacity available to the XenServer host.

Since the VM has new virtual network hardware, the MAC addresses it sees will also be different. Follow the Linux cloning guidelines (see the section called “Preparing to clone a Linux VM”) for customizing the configuration files to make the VM re-run any hardware detection scripts at startup.

General Guidelines for Virtualizing Physical Servers

When considering how to best begin virtualizing a collection of physical servers, it is best to gain some comfort level and experience with virtualizing servers that are more simply configured, moving later to servers with more complex configurations.

Good candidates typically include servers that are used for test and development environments, and servers used for in-house IT infrastructure (intranet web servers, DNS, NIS, and other network services, etc.). Typically servers that are doing heavily CPU-intensive tasks (sophisticated mathematical modeling, video rendering) or are I/O-intensive (high-traffic commercial web sites, highly-used database servers, streaming audio/video servers) are not the best candidates for virtualization at the start.

Once you've identified some physical servers that seem reasonable to work on first, you should take a close look at how you are currently using them. What applications are they hosting? How I/O intensive are they? How CPU-intensive are they?

To make a reasonable assessment, you should gather a reasonable amount of data on the current physical servers that you are thinking about virtualizing. Look at system monitoring data for disk usage, CPU usage, memory usage, and network traffic, and consider both peak and average values.

Good candidates for virtualization are:

servers whose CPU and memory usage and NIC and disk throughput are low will be more likely to coexist as a VM on a XenServer Host with a few other VMs without unduly constraining its performance.
servers that are a few years old - so their performance as VMs hosted on a newer server would be comparable to their existing state.
servers that don't use any incompatible hardware which cannot be virtualized, such as dongles, serial or parallel ports, or other unsupported PCI cards (serial cards, crypto accelerators, etc.).

Once you have identified a set of machines that you want to virtualize, you should plan the process to accomplish the task. First, provision the physical servers that will serve as your XenServer Hosts. The chief constraint on the number of VMs you can run per XenServer Host is system memory.

Next, plan how you will create the VMs. Your choices are to P2V an existing server, install a fresh server from network-mounted vendor media, or install a base operating system using a pre-existing template.

If you P2V an existing server, it's best to P2V a test instance of the server, and run it in parallel with the existing physical server until you are satisfied that everything works properly in the virtual environment before re-purposing the existing physical machine.

Next, plan how to arrange the desired VMs on the XenServer Hosts. Don't "mix up" servers - assign VMs to specific XenServer Hosts, giving consideration to complementary resource consumption (mixing CPU-intensive and I/O-intensive workloads) and complementary peak usage patterns (for instance, assigning overnight batch processing and daytime interactive workloads to the same XenServer Host).

For configuring individual VMs themselves, keep these guidelines in mind:

create single-processor VMs unless you are serving a multi-threaded application that will perform demonstrably better with a second virtual CPU.
when you configure the memory settings for a VM, consult the documentation for the guest operating system you plan to run in that VM and for the applications you plan to run on them.

Cloning an existing VM

You can make a copy of an existing VM by cloning from a template. Templates are just ordinary VMs which are intended to be used as master copies to instantiate copies from. A VM can be customized and converted into a template, but be sure to follow the appropriate preparation procedure for the VM (see the section called “Preparing to clone a Windows VM” for Windows and the section called “Preparing to clone a Linux VM” for Linux). Templates cannot be used as normal VMs without first cloning them.

XenServer has two mechanisms for cloning VMs: a full copy, or a faster "Copy-on-Write" (CoW) mode which only writes modified blocks to disk. The CoW mode is only supported for file-backed VMs. CoW is designed to save disk space and allow fast clones, but will slightly slow down normal disk performance. A template can be fast-cloned multiple times without slowdown, but if a template is cloned into a VM and the clone converted back into a template, disk performance can linearly decrease depending on the number of times this has happened. In this event, the vm-copy CLI command can be used to perform a full copy of the disks and restore expected levels of disk performance.

Resource pools introduce some complexities around creating custom templates and cloning them. If you create a template on a server in the pool, and all virtual disks of the source VM are on shared storage repositories (SR), the operation of cloning that template will be forwarded to any server in the pool that can see those shared SRs. However, if you create the template from a source VM that has any virtual disks on a local SR, then the clone operation can only execute on the server that can see this SR.

Importing an exported VM

You can make a VM by importing an existing exported VM. Like cloning, exporting and importing a VM is a means for creating additional VMs of a certain configuration. You might, for example, have a special-purpose server configuration that you use many times. Once you have set up a VM the way you want it, you can export it, and import it later at any time you want to create another copy of your specially-configured VM. Export and import also provides a way to move a VM to another XenServer Host that is not part of a resource pool.

When importing a VM, you can choose to preserve the MAC address on any virtual network interfaces associated with it. If you do choose to generate a new MAC address, be sure to follow the appropriate preparation procedure for the imported VM (see the section called “Preparing to clone a Windows VM” for Windows and the section called “Preparing to clone a Linux VM” for Linux).

Importing an exported VM will take some time, and depends on the size of the VM and the speed and bandwidth of the network connection between the XenServer Host and XenCenter.

Note

Note that an exported VM that originated on one XenServer Host might or might not be able to be resumed on a different XenServer Host. For example, if a Windows VM created on a XenServer Host with an Intel VT-enabled CPU is exported, then imported to a XenServer Host with an AMD-V CPU, it will not start.

Exporting a VM

An existing VM can be exported via XenCenter or via the CLI. This section describes using the CLI. For details on exporting using XenCenter, see the XenCenter online Help.

The following procedure assumes that you have multiple XenServer Hosts and that you are administering them using the CLI on a separate machine (that is, a machine that is not one of the XenServer Hosts) where you can maintain a library of export files. Avoid exporting a VM to a XenServer Host filesystem.

To export a VM using the CLI

If it is running, shut down VM that you want to export.
Export the VM:
```
xe vm-export -h <hostname> -u <root> -pw <password> vm=<VM name> \
filename=<pathname of file>
					
```
Note
Be sure to include the .xva extension when specifying the export filename. If the exported VM does not have this extension and you attempt to import it via XenCenter, it will fail to recognize the file as a valid XVA file.
The export process will probably take some time. When finished, the command prompt returns.

Importing a VM

An existing exported VM file can be imported via XenCenter or via the CLI. This section describes using the CLI. For details on importing using XenCenter, see the XenCenter online Help.

The following procedure assumes that you are administering the XenServer Host using the CLI on a separate machine (that is, a machine that is not one of your XenServer Hosts) where you maintain a library of export files.

To import a VM using the CLI

To import the VM to the default SR on the target XenServer Host:

xe vm-import -h <hostname> -u <root> -pw <password> \
filename=<pathname of export file>

You can import the VM to another SR on the target XenServer Host by adding the optional sr-uuid parameter:

xe vm-import -h <hostname> -u <root> -pw <password> \
filename=<pathname of export file> sr-uuid=<UUID of target SR>

You can also preserve the MAC address of the original VM by adding the optional preserve set to true:

xe vm-import -h <hostname> -u <root> -pw <password> \
filename=<pathname of export file> preserve=true

The import process will probably take some time. When finished, the command prompt returns the UUID of the newly-imported VM.

Installing Windows VMs

Table of Contents

Making the ISO available to XenServer Hosts

Copying ISOs to local storage

Windows paravirtualized drivers

Remote Desktop

Preparing to clone a Windows VM

Release Notes

General Windows Issues
Windows 2003 Server
Windows XP SP2
Windows 2000 Server
Windows Vista

XenServer allows you to install Windows 2000 SP4, Windows Server 2003 (32-/64- bit), or Windows XP SP2 into a VM. Installing Windows VMs on XenServer Host requires hardware virtualization support (Intel VT or AMD-V).

Installing a Windows VM can be broken down into two main steps:

installing the Windows operating system
installing the paravirtualized device drivers

Windows VMs are installed by cloning an appropriate template from either XenCenter or the CLI. The templates for individual guests have predefined platform flags set which define the configuration of the virtual hardware. For example, all Windows VMs are installed with the ACPI Hardware Abstraction Layer (HAL). If you subsequently change one of these VMs to have multiple virtual CPUs, Windows automatically switches the HAL to multi-processor mode.

The available Windows templates are:

Windows Vista: can be used to install Windows Vista 32-bit. The Enterprise edition is supported.
Windows Server 2003: can be used to install Windows Server 2003 32-bit SP0, SP1, and SP2. The Server, Enterprise, Data Centre, and SBS editions are supported.
Windows Server 2003 x64: can be used to install Windows Server 2003 64-bit. The Server, Enterprise, Data Centre, and SBS editions are supported.
Windows 2000 SP4: can be used to install Windows 2000 Server Service Pack 4. Earlier service packs are not supported.
Windows XP SP2: can be used to install Windows XP Service Pack 2. Earlier service packs are not supported.

The Windows VM can be installed either from a install CD in a physical CD-ROM on the XenServer Host, or from an ISO image of your Windows media (see Appendix A, Creating ISO images for information on how to make an ISO image from a Windows install CD and make it available for use).

Making the ISO available to XenServer Hosts

To make an ISO library available to XenServer Hosts, create an external NFS or SMB/CIFS share directory. The NFS or SMB/CIFS server must be set to allow root access to the share. For NFS shares, this is accomplished by setting the no_root_squash flag when you create the share entry in /etc/exports on the NFS server.

Then either use XenCenter to attach the ISO library, or connect to the host console and type in:

xe-mount-iso-sr host:/volume

Additional arguments to the mount command may be passed in, for advanced use.

If making a Windows SMB/CIFS share available to the XenServer host, either use XenCenter to make it available, or connect to the host console and type in:

xe-mount-iso-sr unc_path -t smbfs -o username=myname/myworkgroup

The unc_path argument should have back-slashes replaces by forward-slashes. -t cifs can be used for CIFS instead of SMB. Examples:

xe-mount-iso-sr //server1/myisos -t cifs -o username=johndoe/mydomain
xe-mount-iso-sr //server2/iso_share -t smbfs -o username=alice

After mounting the share, any ISOs in it should be available by name from the CD pulldown list in XenCenter, or as CD images from the CLI commands. The ISO should be attached to an appropriate Windows template:

Windows Server 2003
Windows Server 2003 x64
Windows 2000 SP4
Windows XP SP2

Copying ISOs to local storage

In XenServer 3.2 and earlier, ISOs could be copied directly to the control domain into the /opt/xensource/packages/iso directory. In XenServer 4.1.0 hosts, this directory is reserved for use of the built-in ISO images, and is not intended for general use. This directory is considered to be identical across hosts in a resource pool, and CD images may fail to attach if the contents are modified.

To use local ISO storage from the control domain

Log onto the host console.
Create a directory to copy the local ISOs into:
```
mkdir -p /var/opt/xen/iso_import
				
```

Create an ISO storage repository by:

xe sr-create name-label=<name> type=iso \
device-config:location=/var/opt/xen/iso_import/<name> \ 
device-config:legacy_mode=true content-type=iso

Copy the ISO images into this directory, taking care not to fill up the control domain filesystem.
Verify that the ISO image is available for use by xe vdi-list, or checking the CD drop-down box in XenCenter.

Warning

Be extremely careful with copying ISOs directly onto the control domain filesystem, as it has limited space available. A network share is a much safer mechanism for storing large numbers of ISO images. If the control domain does fill up, unpredictable behavior will result.

Windows paravirtualized drivers

The Citrix paravirtualizated network and SCSI drivers provide high performance I/O services without the overhead of traditional device emulation found in first-generation virtualization products. During the installation of a Windows operating system, Xen will use traditional device emulation to present a standard IDE controller and a standard network card to the Virtual Machine. This allows Windows to complete its installation using built-in drivers, but with reduced performance due to the overhead inherent in emulation of the controller drivers.

After Windows is installed, you install the Citrix high-speed paravirtualized drivers. These are on an ISO available to the virtual CD-ROM drive of the Virtual Machine. These drivers replace the emulated devices and provide high-speed transport between Windows and the XenServer product family software.

Note

While a Windows VM will function without them, performance is significantly hampered unless these drivers are installed. Running Windows VMs without these drivers is not supported. Some features, such as live relocation across physical hosts, will only work with the paravirtual drivers installed and active.

The Windows paravirtualized drivers ISO can be attached to the VM by using the Install Tools menu in XenCenter, or by directly inserting the built-in xs-tools.iso ISO image to the VM using the CLI. Once the ISO is attached, double-click on the xensetup.exe installer executable and follow the on-screen prompts.

Normally, the Windows paravirtualized drivers are installed by default in the directory C:\Program Files\XenSource\Drivers on the VM.

Remote Desktop

The graphical console for Windows can be either a standard console via emulated graphics card, or an RDP connection.

For Windows VMs, there is a Switch to Remote Desktop button on the Console tab. Clicking it disables the standard graphical console, and switches to using Remote Desktop instead.

The button will be grayed out if you do not have Remote Desktop enabled in the VM, and the paravirtualized drivers must be installed.

To enable Remote Desktop on a Windows VM

From the Start menu, select Control Panel.
From the Control Panel window, select System.
In the System Properties dialog box, select the Remote tab.
In the Remote Desktop section of this dialog box, check the checkbox labeled Allow users to connect remotely to this computer (Windows XP) or Enable Remote Desktop on this computer (Windows 2003 Server).
If you want to select any non-administrator users that can connect to this Windows VM, click the Select Remote Users... button and provide the usernames. (Users with Administrator privileges on the Windows domain can connect by default.)

Preparing to clone a Windows VM

You need to use the Windows utility sysprep to prepare a Windows VM for cloning. This is the only supported way to properly clone a Windows VM.

Computers running Windows operating systems use a Security ID (SID) to uniquely identify themselves. When cloning a Windows VM, it is important to take steps to ensure the uniqueness of these Security IDs. Cloning an installation without taking the recommended system preparation steps can lead to duplicate SIDs and other problems. Because the SID identifies the computer or domain as well as the user, it is critical that it is unique. Refer to the Microsoft KnowledgeBase article 162001, "Do not disk duplicate installed versions of Windows," for more information.

Sysprep modifies the local computer Security ID (SID)to make it unique to each computer. The Sysprep binaries are on the Windows product CDs in the \support\tools\deploy.cab file.

Here are the overall steps you need to follow to clone Windows VMs:

Create, install, and configure the Windows VM as desired.
Apply all relevant Service Packs and updates.
Install the Citrix PV drivers.
Install any applications and perform any other tailoring that is desired.
Copy the contents of \support\tools\deploy.cab from the Windows product CD to a new \sysprep folder in the VM.
Run sysprep (this will shut down the VM when it completes).
In XenCenter, convert the VM into a template.
Clone the newly created template into new VMs as required.
When the cloned VM starts, it will get a new system ID and name, then run a mini-setup to prompt for configuration values as necessary, and finally restart, before being available for use.

Note

The original, sysprepped VM (the "source" VM) should not be restarted again after the sysprep stage, and should be converted to a template immediately afterwards to prevent this. If the source VM is restarted, sysprep must be run on it again before it can be safely used to make additional clones.

For more information on using sysprep, refer to the Microsoft TechNet page "Windows System Preparation Tool."

Release Notes

There are many versions and variations of Windows with different levels of support for the features provided by XenServer. This section lists notes and errata for the known differences.

Installing Linux VMs

XenServer supports the installation of many Linux distributions into paravirtualized VMs. There are four installation mechanisms at present: complete distributions provided as built-in templates, Physical-to-Virtual (P2V) conversion of an existing native instance (see the section called “Physical to Virtual Conversion (P2V)”), using the vendor media in the server's physlca DVD/CD drive, and using the vendor media to perform a network installation.

Installing Linux VMs requires the Linux Pack to be installed onto the XenServer Host.

Warning

If you have not installed the Linux Pack, and you are using XenCenter to install VMs, the New VM wizard will show only Windows choices in the list. You might be tempted to select Other install media to install a Linux VMs. This will not work properly and is not supported.

The Other install media template is meant for advanced users who want to attempt installing VMs running other unsupported operating systems. XenServer has been tested running only the supported distributions and specific versions covered by the standard supplied templates, and any VMs installed using the Other install media template are not supported.

The supported Linux distributions are:

Distribution	Built-in	P2V	Vendor Install from CD	Vendor Install from network repository
Debian Sarge 3.1	X
Debian Etch 4.0	X
Red Hat Enterprise Linux 3.6-3.8		X
Red Hat Enterprise Linux 4.1-4.5		X		X
Red Hat Enterprise Linux 5.0-5.1 32-bit			X	X
Red Hat Enterprise Linux 5.0-5.1 64-bit			X	X
SUSE Linux Enterprise Server 9		X
SUSE Linux Enterprise Server 10 SP1				X
CentOS 4.5-4.6			X	X
CentOS 5.0-5.1 32-bit			X	X
CentOS 5.0-5.1 64-bit			X	X
Oracle Enterprise Linux 5.0-5.1 32-bit			X	X
Oracle Enterprise Linux 5.0-5.1 64-bit			X	X

Note

Distributions which use the same installation mechanism as Red Hat Enterprise Linux 5 (e.g. Fedora Core 6) might be successfully installed using the same template. However, distributions not present in the above list are not supported.

Installation of a built-in distribution

This is the simplest way of installing a VM. The template provided with XenServer can be used to directly create a VM running version 3.1 (Sarge) or 4.0 (Etch) of the Debian Linux distribution without need for vendor installation media and without performing a P2V conversion of an existing physical server.

The VMs are instantiated by using the vm-install from the CLI, or by cloning the template using XenCenter. For example, using the CLI on Linux:

# xe vm-install template=Debian\ Etch\ 4.0 new-name-label=ExampleVM
f21cd819-5b7d-002d-7a1e-861a954e770

When the VM is first booted, it will prompt you for a root password, a VNC password (for graphical use), and a hostname. After values are entered for these, it will finish at a standard login prompt, ready for use. You will need to add a network interface if installed via the CLI.

Installing Linux from vendor media to a VM

This version of XenServer supports installation of the following Linux operating systems from vendor media in the XenServer Host DVD/CD-ROM drive:

Red Hat Enterprise Linux 5.0-5.1, 32-bit
Red Hat Enterprise Linux 5.0-5.1, 64-bit
CentOS 4.5-4.6
CentOS 5.0-5.1, 32-bit
CentOS 5.0-5.1, 64-bit
Oracle Enterprise Linux 5.0-5.1, 32-bit
Oracle Enterprise Linux 5.0-5.1, 64-bit

Other Linux operating systems need to be installed from a network installation server. See the section called “Installing Linux from a network installation server to a VM”.

To install a supported Linux VM from vendor media via the CLI

Insert the vendor installation CD into the XenServer Host's CD drive.
Enter the command
```
# xe template-list
			
```
to find the name of the template corresponding to the OS you want to install.

Enter the command

xe vm-install template="<template name>" new-name-label=<name for VM>

This returns the UUID of the new VM.

Get the UUID of the new VM's root disk:

xe vbd-list vm-uuid=<VM UUID> params=uuid --minimal

Using the UUID returned, set the root disk to be not be bootable:
```
xe vbd-param-set uuid=<root disk UUID> bootable=true 
			
```
Get the name of the physical CD drive on the XenServer Host:
```
# xe cd-list
			
```
The result of this command should give you something like SCSI 0:0:0:0 for the name-label field.
Add a virtual CD-ROM to the new VM using the XenServer Host's CD drive name-label in the cd-name argument:
```
xe vm-cd-add uuid=<VM UUID> cd-name="<host CD drive name label>" device=3
			
```

Get the UUID of the VBD corresponding to the new virtual CD drive:

# xe vbd-list vm-uuid=<VM UUID> type=CD params=uuid --minimal

Make the virtual CD's VBD bootable:

# xe vbd-param-set uuid=<VM UUID> bootable=true

Set the VM's install repository to be the CD drive:

# xe vm-param-set uuid=<VM UUID> other-config:install-repository="cdrom"

Start the VM:
```
xe vm-start uuid=<VM UUID>
			
```
Open a text console to the VM with XenCenter or an SSH terminal and follow the steps to perform the OS installation.

Note

The console in XenCenter does not support VNC during the OS installation. However, it is possible to perform a graphical installation rather than the text-based installation if you connect an external VNC client to the new VM. You must also set a couple of additional other-config parameters for the VM before starting it. The command to set these is

xe vm-param-set uuid=<VM UUID> other-config:install-vnc=1 \
other-config:install-vncpasswd=<VNC password>

Installing Linux from a network installation server to a VM

The XenServer guest installer allows you to install an operating system from a network-accessible copy of vendor media into a VM. In preparation for installing from vendor media, you need to make an exploded network repository of your vendor media (not ISO images), exported via NFS, HTTP or FTP accessible to the XenServer Host administration interface. See Appendix B, Setting Up a Red Hat Installation Server for information on how to copy a set of installation CDs to a network drive. For further information, see the section "Preparing for a Network Installation" in the Red Hat Enterprise Linux 4 Installation Guide for details.

The network repository must be accessible from the control domain of the XenServer host, normally via the management interface. The URL should point to the base of the CD/DVD image on the network server, and be of the form:

HTTP: http://<server>/<path>
FTP: ftp://<server>/<path>
NFS: nfs:<server>:/<path>

The XenServer New VM Wizard provides an additional step for vendor-installable templates which prompts for the repository URL. When using the CLI, install the template as normal using vm-install and then set the other-config-install-repository key to the value of the URL. When the VM is subsequently started, it will begin the network installation process.

Note

When installing a new Linux-based VM, it is important to fully finish the installation and reboot it before performing any other operations on it. This is analogous to not interrupting a Windows installation, which would leave you with a non-functional VM.

To install a Linux VM from a network-accessible copy of vendor media via the CLI

Issue the command

xe vm-install template=<template> new-name-label=<name for VM>

This returns the UUID of the new VM.

Find the UUID of the network that you want to connect to. For example, if it's the one attached to xenbr0:
```
xe network-list bridge=xenbr0 --minimal
			
```

Create a VIF to connect the new VM to this network:

xe vif-create vm-uuid=<VM UUID> network-uuid=<network UUID> mac=random device=0

Set the install-repository key of the other-config parameter to the path of your network repository. For example, we'll use http://server/RedHat/5.0 as the URL of the vendor media:
```
xe vm-param-set uuid=<VM UUID> \
other-config:install-repository=<http://server/RedHat/5.0>
			
```
Start the VM:
```
xe vm-start uuid=<VM UUID>
			
```
Connect to the VM console via XenCenter or VNC and perform the OS installation.

Physical-to-Virtual Installation of a Linux VM

Older Linux distributions such as Red Hat Linux Enterprise 3.6 do not support Xen directly, and are typically legacy installations which benefit from virtualization for the purposes of server consolidation or hardware upgrades. The XenServer P2V feature analyzes existing installations and converts them into VMs.

When an installation is converted into a VM using P2V (see the section called “Physical to Virtual Conversion (P2V)”), the kernel used is also automatically switched to a Xen paravirtualized kernel. XenServer contains ports of the Red Hat Enterprise Linux 3/4 and SUSE Enterprise Linux 9 kernels to support the native Xen hypervisor interface directly. These kernels are present in the built-in xs-tools.iso image in the default CD list, or via the Install XenServer Tools command in the VM menu in XenCenter.

Warning

While a VM is in the process of being installed via P2V, do not attempt to perform any operations on it.

Guest Installation Network

During the installation of a VM via P2V, a special network is used to assign a temporary IP address to the VM to enable the installation to proceed. It is possible that the range of IP addresses used might conflict with real IP addresses already in use in your network. The default range of IP addresses is 192.168.128.1 to 192.168.128.254, and the default netmask is 255.255.255.0.

To change the guest installer network values

Open a text console on the XenServer Host or install the CLI for remote use.
Find the guest installer network:
```
xe network-list
				
```
The command will return the list of networks available to the XenServer Host. The one you want has the name-label Guest installer network.
Examine the other-config parameters of the guest installer network:
```
xe network-param-list uuid=<UUID of the guest installer network>
```
The command will a subset of the guest installer network's parameters, including the other-config parameter. If the values are set to the default described above, you will see the line:
```
other-config (MRW): is_guest_installer_network: true; ip_begin: 169.254.0.1; \
ip_end: 169.254.255.254; netmask: 255.255.0.0
				
```

To change the IP address range the guest installer network will use, edit the ip_begin, ip_end, and netmask values as follows:

xe network-param-set uuid=<UUID of the guest installer network> \
other-config:ip_begin=<desired IP range beginning> \
other-config:ip_end=<desired IP range end> \
other-config:netmask=<desired netmask>

Do not change the value of the parameter is_guest_installer_network.

Installing the Linux guest agent

Although all the supported Linux distributions are natively paravirtualized (and thus do not need special drivers for full performance), XenServer includes a guest agent which provides additional information about the VM to the host. This additional information includes:

Linux distribution name and version (major, minor revision).
Kernel version (uname).
IP address of each Ethernet interface.
Total and free memory within the VM.

It is important to install this agent and keep it up-to-date (see Updating VMs) as you upgrade your XenServer host.

To install the guest agent

The files required are present on the built-in xs-tools.iso CD image, or alternatively by using the “Install Tools” option in XenCenter.
Mount the image into the guest via:
```
mount /dev/xvdd /mnt
			
```
Execute the installation script as the root user:
```
/mnt/Linux/install.sh
			
```
If the VM was upgraded from XenServer 3.1, you will also need to run:
```
chkconfig xengmond off
			
```
If the kernel has been upgraded, or the VM was upgraded from a previous version, reboot the VM now.

Note

CD-ROM drives and ISOs attached to Linux Virtual Machines appear as /dev/xvdd instead of as /dev/cdrom as you might reasonably expect. This is because they are not "true" CD-ROM devices, but normal devices. When the CD is ejected by either XenCenter or the CLI, it hot-unplugs the device from the VM and the device disappears. This is different from Windows Virtual Machines, where the CD remains in the VM in an empty state.

Preparing to clone a Linux VM

When a Linux VM is cloned, some virtual hardware parameters are changed in the new VM. The VM may need to be customized in order to be aware of these changes. For instructions for specific supported Linux distributions, please see the section called “Release Notes”.

Machine Name

Of course, a cloned VM is another computer, and like any new computer in a network, it must have a unique name within the network domain it is part of.

IP address

A cloned VM must have a unique IP address within the network domain it is part of. This is not a problem in general if DHCP is used to assign addresses; when the VM boots the DHCP server will assign it an IP address. If the cloned VM had a static IP address, the clone must be given an unused IP address before being booted.

MAC address

In some cases, the MAC address of a cloned VM's virtual network interface is recorded in the network configuration files. After the VM is cloned, the new cloned VM has a different MAC address. Therefore, when started, the network does not come up automatically.

Some Linux distributions use udev rules to remember the MAC address of each network interface, and persist a name for that interface. This is intended so that the same physical NIC always maps to the same ethn interface, which is particularly useful with removable NICs (like laptops). But this behavior is problematic in the context of Virtual Machines. For example, if you configure two virtual NICs when you install a VM, and then shut it down and remove the first NIC, on reboot XenCenter shows just one NIC, but calls it eth0. Meanwhile the VM is deliberately forcing this to be eth1. The result is that networking doesn't work.

If the VM uses persistent names, the best thing to do is to turn these rules off. If for some reason you do not want to turn persistent names off, be aware that you will need to reconfigure networking inside the VM in the usual way, and the information shown in XenCenter will be out of sync with reality.

Time handling in Linux VMs

By default, the clocks in a Linux VM are synchronized to the clock running on the control domain, and cannot be independently changed. This mode is a convenient default, since only the control domain needs to be running the NTP service to keep accurate time across all VMs. Upon installation of a new Linux VM, make sure you change the time-zone from the default UTC to your local value (see the section called “Release Notes” for specific distribution instructions).

Individual Linux VMs can be set to maintain independent times

From a root prompt on the VM, type the command: echo 1 > /proc/sys/xen/independent_wallclock
This can be persisted across reboots by changing the /etc/sysctl.conf configuration file and adding:
```
# Set independent wall clock time
xen.independent_wallclock=1
```
As a third alternative, the independent_wallclock=1 may also be passed as a boot parameter to the VM.

Configuring VNC for VMs

With the exception of VMs based on the Debian templates, VMs might not be set up to support VNC by default. For example, if you P2V a server that does not have a VNC server installed, the resulting VM won't have VNC installed either. Before you can connect with the XenCenter graphical console, you need to ensure that the VNC server and an X display manager are installed on the VM and properly configured. This section describes the procedures for configuring VNC on each of the supported Linux operating system distributions to allow proper interactions with the XenCenter graphical console.

CentOS-based VMs should use the instructions for the Red Hat-based VMs below, as they use the same base code to provide graphical VNC access. CentOS 4 is based on Red Hat Enterprise Linux 4, and CentOS 5 is based on Red Hat Enterprise Linux 5.

Setting up Red Hat-based VMs for VNC

Note

Before setting up your Red Hat VMs for VNC, be sure that you have installed the Linux guest agent. See the section called “Installing the Linux guest agent” for details.

In order to configure VNC on Red Hat VMs, you need to modify the GDM configuration. The GDM configuration is held in a file whose location varies depending on the version of Red Hat Linux you are using. Before modifying it, we first must determine the location of this configuration file; this file will then be modified in a number of subsequent procedures in this section.

Determining the location of your VNC configuration file

If you are using Red Hat Linux version 3 or 4 the GDM configuration file is /etc/X11/gdm/gdm.conf. This is a unified configuration file that contains both default values as specified by the provider of your version of GDM in addition to your own customized configuration. This type of file is used by default in older versions of GDM, as included in these version of Red Hat Linux.

If you are using Red Hat Linux version 5 the GDM configuration file is /etc/gdm/custom.conf. This is a split configuration file that contains only user-specified values that override the default configuration. This type of file is used by default in newer versions of GDM, as included in these versions of Red Hat Linux.

Configuring GDM to use VNC

As root at the prompt in the VM's text console, type in rpm -q vnc-server gdm. The package names vnc-server and gdm should appear, with their version numbers specified.If these package names are displayed, the appropriate packages are already installed. If you see a message saying that one of the packages is not installed, then you may not have selected the graphical desktop options during installation. You will need to install these packages before you can continue. See the appropriate Red Hat Linux x86 Installation Guide for details regarding installing additional software on your VM.
Open the GDM configuration file with your preferred text editor and add the following lines to the file:
```
[server-VNC]
name=VNC Server
command=/usr/bin/Xvnc -SecurityTypes None -geometry 1024x768 -depth 16 \
-BlacklistTimeout 0
flexible=true
```
- With configuration files as found on Red Hat Linux 3 and 4, this should be added above the [server-Standard section.
- With configuration files as found on Red Hat Linux 5, this should be added into the empty [servers] section.
Modify the configuration so that the Xvnc server is used instead of the standard X server:
- If you are using Red Hat Linux 3 or 4, there will be a line just above that says:
```
0=Standard
```
  Modify it to read:
```
0=VNC
```
- If you are using Red Hat Linux 5 or greater, you will need to add the above line just below the [servers] section and before the [server-VNC] section.
Save and close the file.

Restart GDM for your change in configuration to take effect, by running /usr/sbin/gdm-restart.

Note that, for Red Hat Linux, runlevel 5 is used for graphical startup. If your installation is configured to start up in runlevel 3, you will need to change this in order for the display manager to be started (and therefore to get access to a graphical console). Please refer to the section called “Checking runlevels” for further details.

Firewall settings

The firewall configuration by default does not allow VNC to traffic to go through. If you have a firewall between the VM and XenCenter, you need to allow traffic over the port that the VNC connection uses. By default, a VNC server listens for connections from a VNC viewer on TCP port 5900 + N, where N is the display number (usually just zero). So a VNC server setup for Display-0 will listen on TCP port 5900, Display-1 is TCP-5901, etc. Consult your firewall documentation to make sure these ports are open.

You might want to further customize your firewall configuration if you want to use IP connection tracking or limit the initiation of connections to be from one side only.

To customize Red Hat-based VMs firewall to open the VNC port

For Red Hat Linux 3, use redhat-config-securitylevel-tui.For Red Hat Linux 4 and 5, use system-config-securitylevel-tui.
Select “Customize” and add 5900 to the other ports list.

Alternatively, you can disable the firewall until the next reboot by using service iptables stop, or permanently by using chkconfig iptables off. This can of course expose additional services to the outside world and reduce the overall security of your VM.

VNC screen resolution

If, after connecting to a Virtual Machine with the graphical console, the screen resolution is mismatched (for example, the VM's display is too big to comfortably fit in the Graphical Console pane), you can control it by setting the VNC server's -geometry parameter as follows:

Open the GDM configuration file with your preferred text editor. Please refer to the section called “Determining the location of your VNC configuration file” for information about determining the location of this file.
Find the [server-VNC] section you added above.
Edit the command line to read, for example,
```
command=/usr/bin/Xvnc -SecurityTypes None -geometry 800x600
```
where the value of the -geometry parameter can be any valid screen width and height.
Save and close the file.

Setting up SLES-based VMs for VNC

Note

Before setting up your SUSE Linux Enterprise Server VMs for VNC, be sure that you have installed the Linux guest agent. See the section called “Installing the Linux guest agent” for details.

SLES has support for enabling “Remote Administration” as a configuration option in YaST. You can select to enable Remote Administration at install time, available on the Network Services screen of the SLES installer. This will allow you to connect an external VNC viewer to your guest to view the graphical console; the methodology for using the SLES remote administration feature is slightly different than that provided by XenCenter, but it is possible to modify the configuration files in your SUSE Linux VM such that it is integrated with the graphical console feature.

Checking for a VNCserver

Before making configuration changes, you should verify that you have a VNC server installed. SUSE ships the tightvnc server by default; this is a suitable VNC server, but you can also use the standard RealVNC distribution if you prefer.

You can check that you have the tightvnc software installed by running the command:

rpm -q tightvnc

Enabling Remote Administration

If Remote Administration was not enabled during installation of the SLES software, you can enable it as follows:

Open a text console on the VM and run the YaST utility:
```
# yast
		
```
Use the arrow keys to select Network Services in the left menu, then Tab to the right menu and use the arrow keys to select Remote Administration. Press Enter.
In the Remote Administration screen, Tab to the Remote Administration Settings section. Use the arrow keys to select Allow Remote Administration and press Enter to place an X in the checkbox.
Tab to the Firewall Settings section. Use the arrow keys to select Open Port in Firewall and press Enter to place an X in the checkbox.
Tab to the Finish button and press Enter.
A messsage box appears telling you that you will need to restart the display manager for your settings to take effect. Press Enter to acknowledge the message.
The original top-level menu of YaST appears. Tab to the Quit button and press Enter.

Modifying the xinetd configuration

After enabling Remote Administration, you need to modify a configuration file if you want to allow XenCenter to connect, or else use a third party VNC client.

Open the file /etc/xinetd.d/vnc in your preferred text editor.The file contains sections like the following:

service vnc1
{
  socket_type = stream
  protocol    = tcp
  wait        = no
  user        = nobody
  server      = /usr/X11R6/bin/Xvnc
  server_args = :42 -inetd -once -query localhost -geometry 1024x768 -depth 16
  type        = UNLISTED
  port        = 5901
}

Edit the port line to read
```
port = 5900
		
```
Save and close the file.
Restart the display manager and xinetd service with the following commands:
```
/etc/init.d/xinetd restart
rcxdm restart
```

SUSE Linux uses runlevel 5 for graphical startup. If your remote desktop does not appear, verify that your VM is configured to start up in runlevel 5. Refer to the section called “Checking runlevels” for details.

Firewall settings

You might want to further customize your firewall configuration if you want to use IP connection tracking or limit the initiation of connections to be from one side only.

To customize SLES-based VMs firewall to open the VNC port

Open a text console on the VM and run the YaST utility:
```
# yast
		
```
Use the arrow keys to select Security and Users in the left menu, then Tab to the right menu and use the arrow keys to select Firewall. Press Enter.
In the Firewall screen, Tab to the Firewall Configuration: Settings section. Use the arrow keys to select the Allowed Services in the left menu.
Tab to the Firewall Configuration: Allowed Services fields on the right. Use the arrow keys to select the Advanced... button (near the bottom right, just above the Next button) and press Enter.
In the Additional Allowed Ports screen, type 5900 in the TCP Ports field. Tab to the OK button and press Enter.
Tab back to the list of screens on the left side and use the arrow keys to select Start-Up. Tab back to the right and Tab to the Save Settings and Restart Firewall Now button and press Enter.
Tab to the Next button and press Enter, then in the Summary screen Tab to the Accept button and press Enter, and finally on the top-level YaST screen Tab to the Quit button and press Enter.
Restart the display manager and xinetd service with the following commands:
```
/etc/init.d/xinetd restart
rcxdm restart
```

Alternatively, you can disable the firewall until the next reboot by using the rcSuSEfirewall2 stop command, or permanently by using YaST. This can of course expose additional services to the outside world and reduce the overall security of your VM.

VNC screen resolution

If, after connecting to a Virtual Machine with the Graphical Console, the screen resolution is mismatched (for example, the VM's display is too big to comfortably fit in the Graphical Console pane), you can control it by setting the VNC server's -geometry parameter as follows:

Open the /etc/xinetd.d/vnc file with your preferred text editor and find the service_vnc1 section (corresponding to displayID 1).
Edit the -geometry argument in the server-args line to the desired display resolution. For example,
```
server_args  = :42 -inetd -once -query localhost -geometry 800x600 -depth 16
```
where the value of the -geometry parameter can be any valid screen width and height.
Save and close the file.

Restart the vnc server:

/etc/init.d/xinetd restart
rcxdm restart

Setting up Debian-based VMs for VNC

The built-in Debian Sarge and Etch templates come pre-configured with VNC setup and ready use. However, the default VNC configuration in Debian does not permit the root administration user to log in by default. To log in by VNC, you can either:

Log in to the text console and create a new, unprivileged user via the adduser command. This is the recommended course of action.
At the graphical console login prompt, select Actions, Configure the Login Manager, type in your root password, then select Security, Allow local system administrator login, and finally select Close.

If you need to reset the VNC password, use the command

vnc4passwd /etc/vncpass

Checking runlevels

Red Hat and SUSE Linux VMs use runlevel 5 for graphical startup. This section describes how to verify that your VM is configured to start up in runlevel 5 and how to change it if it is not.

Check /etc/inittab to see what the default runlevel is set to. Look for the line that reads:
```
id:n:initdefault:
```
If n is not 5, edit the file to make it so.
You can run the command telinit q ; telinit 5 after this change to avoid having to actually reboot to switch runlevels.

Release Notes

Most modern Linux distributions support Xen paravirtualization directly, but have different installation mechanisms and some kernel limitations.

Debian Sarge 3.1 and Etch 4.0

XenServer includes a custom Xen kernel for Debian guests installed via the built-in template. This kernel is a cut-down version of the dom0 kernel used in XenServer, and as such is the most heavily tested and reliable guest kernel available.

When a Debian VM is first booted, it will have a custom script which will prompt for details such as hostname and root passwords. Since these are executed at a low run-level, it will prevent a freshly installed Debian guest from rebooting until the requested information is entered. In order to bypass the first-boot scripts and boot non-interactively, you must pass the noninteractive flag to the kernel arguments.

After installation, the time-zone in a Debian VM defaults to UTC (see the section called “Time handling in Linux VMs”). It can be changed to your local value by using the tzconfig command.

To prepare a Debian guest for cloning (see the section called “MAC address”), Ethernet name persistence must be disabled. For Debian Sarge VMs, name persistence is controlled through /etc/udev/persistent-net-generator.rules in the udev package. It can be disabled by removing the following symlink to that file by:

rm -f /etc/udev/rules.d/z45_persistent-net-generator.rules

Red Hat Enterprise Linux 3

XenServer includes a custom port of the RHEL3.8 kernel with native Xen paravirtualized guest support. This kernel is installed during the P2V process for RHEL3.6-3.8 guests. Since the kernel is based on Linux 2.4, the following limitations apply:

Only 3 virtual network interfaces are supported.
VMs with multiple VCPUs cannot be suspended. If you wish to suspend these guests, you must reduce the number of VCPUs to 1 while the guest is halted.

Before performing a P2V conversion from an existing RHEL3 installation, ensure that the /etc/fstab file in the guest contains an entry for the /boot mount point. This partition contains the files which are changed by the P2V process to give the resulting VM a para-virtual kernel.

Red Hat Enterprise Linux 4

XenServer includes a custom port of the RHEL 4.6 kernel with additional bug fixes and expanded Xen support. This kernel is used in the vendor installation templates of RHEL 4.1 and 4.4, but not in the RHEL 4.5 or 4.6 templates (since RHEL 4.5 is the first release with native Xen support).

The issues below have been reported upstream to Red Hat and are already fixed in our kernel (which can be installed by using the /mnt/Linux/install.sh script in the built-in xs-tools.iso CD image):

During the resume operation on a suspended VM, allocations can be made that can cause swap activity which cannot be performed because the swap disk is still being reattached. (Red Hat Bugzilla 429103).)
The netfront driver in the RHEL 4.6 kernel has issues with the iptables firewall due to the use of checksum offloading. To work around this issue, disable checksum offload on the VIF associated with the device in the control domain of the XenServer Host on which your RHEL 4.6 VM runs. First determine the UUID of the VIF, by:
```
xe vif-list vm-name-label=examplevm
```
Then disable checksum offload on the VIF by:
```
xe vif-param-set uuid=vif_uuid other_config:ethtool-tx=off
				
```
Only 3 virtual network interfaces are supported.
The 2.6.9-67.ELxen kernel can occasionally enter tickless mode when an RCU is pending. When this triggers, it is usually in synchronize_kernel() which means the guest essentially hangs until some external event (such as a SysRQ) releases it (Red Hat Bugzilla 427998)
Occasional kernel crash on boot in queue_work() (Red Hat Bugzilla 246586)
Disks sometimes do not attach correctly on boot (Red Hat Bugzilla 247265)
Live migration can occasionally crash the kernel under low memory conditions (Red Hat Bugzilla 249867)
Guest kernel can occasionally hang due to other xenstore activity (Red Hat Bugzilla 250381)
If you attempt to install RHEL 4.x on a VM that has more than 2 virtual CPUs (which RHEL 4.x does not support), an error message incorrectly reports the number of CPUs detected.

To prepare a RHEL4 guest for cloning (see the section called “MAC address”), edit /etc/sysconfig/network-scripts/ifcfg-eth0 before converting the VM into a template and remove the HWADDR line. Note that Red Hat recommend the use of Kickstart to perform automated installations, instead of directly cloning disk images (see Red Hat KB Article 2415).

Preparing a RHEL 4.x guest for P2V

Before performing a P2V conversion from an existing RHEL4 installation, ensure that the /etc/fstab file in the guest contains an entry for the /boot mount point. This partition contains the files which are changed by the P2V process to give the resulting VM a para-virtual kernel.

After a successful P2V, some modifications may be needed in older Red Hat Linux 4.x distributions. In order to get LVM working on xvd* devices, you should add the following line under the 'devices {' line in /etc/lvm/lvm.conf:

types = ["xvd", 16]

Red Hat Enterprise Linux 5

XenServer uses the standard Red Hat kernel supplied with RHEL5 as the guest kernel. Any bugs found in this kernel are reported upstream to Red Hat, and are listed below:

During the resume operation on a suspended VM, allocations can be made that can cause swap activity which cannot be performed because the swap disk is still being reattached. (Red Hat Bugzilla 429102).)
After resuming a suspended VM, it might crash with the message kernel BUG at mm/rmap.c:590! (Red Hat Bugzilla 294811)
Only 3 virtual network interfaces are supported.
Random segmentation faults on loading ELF binaries (Red Hat Bugzilla 247261)
Disks sometimes do not attach correctly on boot (Red Hat Bugzilla 247265). This has been fixed in Red Hat Enterprise Linux 5.1.
Soft lockup messages after suspend/resume or live migration (Red Hat Bugzilla 250994). These messages are harmless, but there may be a period of inactivity in the guest during live migration as a result of the lockup.
Network blackout during live relocation for up to a minute (Red Hat Bugzilla 251527). After migration has completed, the kernel sends a gratuitous ARP to cause ARP caches to get refreshed and minimise network downtime. However, carrier detect is delayed in the kernel and so there is a network blackout until the ARP caches expire or the guest generates an ARP for some other reason.

When you install the XenServer xe-guest-utilities RPM, it adds an entry to the yum configuration, allowing you to pick up kernel updates provided by Citrix as they become available.

CentOS 4

Please refer to the section called “Red Hat Enterprise Linux 4” for the list of CentOS 4 release notes.

Unlike RHEL4, CentOS includes a third-party updates mechanism known as yum. The xe-guest-utilities RPM will install a XenServer entry for yum, allowing you to pick up kernel updates provided by Citrix via the standard update mechanism as they become available.

CentOS 5

Please refer to the section called “Red Hat Enterprise Linux 5” for the list of CentOS 5 release notes.

SUSE Enterprise Linux 9

XenServer includes a custom port of the SLES9 SP3 kernel with additional bug fixes and expanded support. This kernel is automatically installed during the P2V process for SLES9 guests.

To prepare a SUSE Linux guest for cloning (see the section called “MAC address”), edit /etc/sysconfig/network/config and edit the line:

FORCE_PERSISTENT_NAMES=yes

FORCE_PERSISTENT_NAMES=no

When you P2V a SLES 9 server, the networking configuration files that were present on the physical server will remain on the VM. You may wish to move these aside, or update them accordingly, when you add virtual interfaces to the VM.

SUSE Enterprise Linux 10 SP1

XenServer uses the standard Novell kernel supplied with SLES 10 SP1 as the guest kernel. Any bugs found in this kernel are reported upstream to Novell and listed below:

Only 3 virtual network interfaces are supported.
Disks sometimes do not attach correctly on boot. (Novell Bugzilla 290346).

Oracle Enterprise Linux 5

Please refer to the section called “Red Hat Enterprise Linux 5” for the list of Oracle Enterprise Linux 5 release notes.

Updating VMs

Table of Contents

Updating Windows operating systems
Updating paravirtualized drivers for Windows VMs
Updating Linux kernels and guest utilities

This chapter discusses updating VMs with new Linux kernel revisions, updating Windows operating systems, applying Windows Service Packs, and updates to XenServer paravirtualized drivers and VM utilities.

Upgrades to VMs are typically required when moving to a new version of XenServer. The following are current issues involving upgrading VMs running on XenServer to this version:

XenMotion of Windows VMs is not supported until the paravirtualized drivers are upgraded.
Suspend/Resume of Windows VMs is not supported until the paravirtualized drivers are upgraded.
The use of certain anti-virus and firewall applications may crash the Windows VM unless the paravirtualized drivers are upgraded.

Updating Windows operating systems

Windows installation disks typically provide an upgrade option if you boot them on a server which has an earlier version of Windows already installed. So if, for example, you have a Windows 2000 server, and you wish to update it to Windows 2003, you can insert the Windows 2003 installation CD in the CD drive and run the setup program to update it.

To uninstall the paravirtualized drivers

Select Control Panel from the Start menu.
In Windows XP, 2000, or 2003, select Add or Remove Programs.In Windows Vista, select Programs, then select Programs and Features.
A list of programs installed on the computer is displayed. Scroll down if necessary and select Citrix XenServer Windows PV drivers Add-on.
In Windows XP, 2000, or 2003, click the Remove button.In Windows Vista, select Uninstall from the toolbar above the list of programs.This will remove the PV drivers add-on. At the end, a message is displayed. Click OK to close the message box.

Once the operating system update is complete, reinstall the PV drivers just as you would after installing a fresh Windows VM. See the section called “Windows paravirtualized drivers” for details.

Updating paravirtualized drivers for Windows VMs

The paravirtualized drivers are present on the built-in xs-tools.iso available to XenCenter via the Install XenServer Tools command from the VM menu, which attaches the CD image containing the drivers to the VM. You can either wait for the auto-run facility, or manually click on the xensetup.exe program. Follow the on-screen prompts to install the new drivers, which will automatically deactivate and upgrade the old drivers.

Updating Linux kernels and guest utilities

The Linux guest utilities can be updated by re-running the Linux/install.sh script from the built-in xs-tools.iso CD image (see the section called “Installing the Linux guest agent”). From time to time, Citrix also supplies updated Linux kernels for supported distributions. Supported distributions are:

Red Hat Enterprise Linux 5.x
CentOS 5.x
Red Hat Enterprise Linux 4.x
CentOS 4.x
Red Hat Enterprise Linux 3.x
Debian Sarge and Etch

The updates are posted online at: http://updates.xensource.com/XenServer/4.1.0/.

For example, the RHEL 3.x kernel would be at: http://updates.xensource.com/XenServer/4.1.0/rhel3x/.

This is of particular importance for RHEL 4.5/4.6 and CentOS 4.5/4.6, where you will get the upstream kernel by default, which has certain limitations (see the section called “Release Notes”).

For yum-enabled distributions (CentOS 4 and 5, RHEL 5), xe-guest-utilities installs a yum configuration file to enable subsequent updates to be done via yum in the standard manner. Note that RHEL 4 in particular does not use yum.

For Debian, /etc/apt/sources.list is populated to enable updates via apt by default.

Appendix A. Creating ISO images

XenServer can use ISO images of CD-ROM or DVD-ROM disks as installation media and data sources for Windows or Linux VMs. This section describes how to make ISOs from CD/DVD media.

On a Linux computer

Put the CD- or DVD-ROM disk into the drive. The disk should not be mounted. To check, type the command:mountIf the disk is mounted, unmount the disk. Refer to your operating system documentation for assistance if required.
As root, type the command
```
dd if=/dev/cdrom of=/path/cdimg_filename.iso
			
```
This will take some time. When the operation is completed successfully, you should see something like
```
1187972+0 records in
1187972+0 records out
			
```
Your ISO file is ready.

On a Windows computer

Windows computers do not have an equivalent operating system command to create an ISO. Most CD-burning tools have a means of saving a CD as an ISO file.One simple and free utility is ISO Recorder. It works on Windows XP SP2, Windows 2000, and Windows Server 2003. Once installed, you simply right-click on a CD/DVD drive and select Create image from CD from the context menu.

Appendix B. Setting Up a Red Hat Installation Server

Table of Contents

Copy installation media

Enable remote access

NFS
FTP
HTTP

This chapter explains how to set up a server as an installation server for Red Hat Linux.

For a server to act as a Red Hat Linux network installation server, you need space on your server to copy the entire contents of each CD onto your server. This is typically the number of CDs or ISO images times 650MB.

Ensure that the space you intend to use is formatted with your chosen filesystem and is mounted. You can check this space with the command:

df -h

Copy installation media

First create a directory to contain the installation files, for example /install
Mount your CD. Refer to your operating system documentation for assistance if needed. This example assumes that it is mounted at /mnt/cdrom:mount /mnt/cdrom
Copy the data from the CD to the installation directory:cp -var /mnt/cdrom/RedHat /install
Unmount the CD:umount /mnt/cdrom
Remove the first CD, put in the next one, and repeat for each of your CDs you have.

Enable remote access

Next, you need to make your installation data available to other machines on the network. You can use NFS, HTTP, or FTP protocols. You can enable all three services on your server or any subset of the three.

NFS

To install over NFS you need to meet certain conditions on the server:

The installation directory must be exportedTo export your installation directory, edit the /etc/exports file and add an entry for /install to it:
```
/install *(ro)
					
```
Save the edited exports file and tell the NFS daemon to reread its configuration file:
```
exportfs -r
					
```
This configures the most basic read-only export to all hosts on our network. If you want to include more advanced options in your export, such as exporting to certain hosts only, or on a certain subnet only, see the man page for the exports file at exports (5).
NFS needs to be installed and runningTo check, type the command:
```
showmount -e hostname
					
```
Entering the showmount command without the hostname parameter will check the local system.If NFS is not active, you will see a message similar to
```
showmount: ServerA: RPC: Program not registered
					
```
Portmap needs to be running, and can be checked with the command:
```
service portmap status
					
```

FTP

To enable installing over FTP, you need to allow FTP access to the installation directory on the server. This can be either anonymous FTP access or access through a named account with a password.

If you want anonymous FTP to point to a different directory, you can use sym links to point to the installation directory on the server.

HTTP

If you have a web server running and want to enable HTTP access to your installation server, then add sym links from your document root to the installation server directory to grant access.

The installation server is now ready to use. Make sure you note the server name or IP address and the directory path to the installation directory you created.

Appendix C. Troubleshooting VM problems

Table of Contents

VM crashes

Linux VMs
Windows VMs

Troubleshooting boot problems on Linux VMs

If you experience odd behavior, application crashes, or have other issues, this chapter is meant to help you solve the problem if possible and, failing that, describes where the application logs are located and other information that can help your XenServer Solution Provider and Citrix track and resolve the issue.

Troubleshooting of installation issues is covered in the XenServer Installation Guide. Troubleshooting of XenServer Host issues is covered in the XenServer Administrator's Guide.

Note

We recommend that you follow the troubleshooting information in this chapter solely under the guidance of your XenServer Solution Provider or Citrix Support.

Citrix provides two forms of support: you can receive free self-help support via the Support site, or you may purchase our Support Services and directly submit requests by filing an online Support Case. Our free web-based resources include product documentation, a Knowledge Base, and discussion forums.

VM crashes

If you are experiencing VM crashes, it's possible that a kernel crash dump might help identify the problem. If the crash is reproducible, follow this procedure to send the dumps to Citrix.

Linux VMs

For Linux VMs, the crashdump behavior can be controlled through the actions-after-crash parameter. The following are the possible values:

Value	Description
preserve	leave the VM in a paused state (for analysis)
coredump_and_restart	record a core dump, then reboot the VM
coredump_and_destroy	record a core dump, leave VM halted
restart	no core dump, just reboot VM (this is the default)
destroy	no coredump, leave VM halted

To enable saving of Linux VM crash dumps

On the XenServer Host, determine the UUID of the desired VM using the command:xe vm-list name-label=<name> params=uuid --minimal
Change the actions-after-crash value using the xe vm-param-set; for example:xe vm-param-set uuid=<vm_uuid> actions-after-crash=coredump_and_restart

Windows VMs

For Windows VMs, the core dump behavior cannot be controlled through the actions-after-crash parameter. By default Windows crash dumps are put into %SystemRoot%\Minidump in the Windows VM itself.

You can configure the VMs dump level by going into My Computer --> Properties --> Advanced --> Startup and Recovery.

Troubleshooting boot problems on Linux VMs

There is a utility script named xe-edit-bootloader in the XenServer Host control domain which can be used to edit the bootloader configuration of a shutdown Linux VM. This can be used to fix problems which are preventing it from booting.

To use:

Use the command
```
xe vm-list
				
```
to ensure that the VM in question is shut down (the value of power-state will be halted).
You can use the UUID as follows:
```
xe-edit-bootloader -u <UUID of Linux VM> -n <partition number>
				
```
or the name-label as follows:
```
xe-edit-bootloader -n <name-label of Linux VM> -n <partition number>
				
```
The partition number represents the slice of the disk which has the filesystem. In the case of the default Debian template, this will be 1 since it is the first partition.
You will be dropped into an editor with the grub.conf file for the specified VM loaded. Make the changes to fix it, and save the file, exit the editor, and start the VM.

Getting Started

Release Notes
Information on known issues, new features and a detailed change-log from the previous version.

HTML PDF

Installation Guide
Read this before commencing a new installation of XenServer, and learn more about maintaining one.

HTML PDF
Manuals
- Virtual Machine Guide
  Instructions on how to use all the supported guest operating systems including Windows and Linux.
  
  HTML PDF
- Reference Manual
  Detailed guide to the product, including storage, networking, and the advanced use of the command-line interface.
  
  HTML PDF
Software Development
- SDK Guide
  An introduction to using the XenServer SDK to develop applications using the XenAPI.
  
  HTML PDF
- API Documentation
  Detailed manual of all the XenAPI data-model components used in the XML-RPC interface.
  
  HTML

XenServer Virtual Machine Installation Guide

Release 4.1.0

Note

Note

Note

Note

Warning

Note

Warning

Note

Note

Note

Getting Started

Manuals

Software Development