8/3/2019 Window CCS and SLES Dual Boot En http://slidepdf.com/reader/full/window-ccs-and-sles-dual-boot-en 1/87 Dual Boot: Windows ® Compute Cluster Server 2003 and SUSE ® Linux Enterprise Server 10 Setup and Configuration Guide Published: July 2007, Updated: December 2007 For the latest information, see http://www.microsoft.com/hpc
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Compute Nodes .................................................................................................................. 6 Windows Compute Cluster Head Node .............................................................................. 6 Infrastructure Server ........................................................................................................... 7
Software Requirements and Versions Used ......................................................................... 7 Windows Compute Cluster Server ..................................................................................... 7 SUSE Linux Enterprise Server ........................................................................................... 7
Hardware Used ...................................................................................................................... 7 2 embedded NC370i Multifunction Gigabit Server Adapters ................................................. 8 Prepare for Windows Compute Cluster Server 2003 and SUSE Linux Enterprise Server 10
SP1 Dual Boot ....................................................................................................................... 8 Prepare for Deployment of Operating Systems ......................................................................... 9
Prepare for SUSE Linux Enterprise Server Deployment ....................................................... 9 Configure the Deployment Server ...................................................................................... 9 Create the AutoYaST Control File .................................................................................... 20 Integrate the Linux HPC Software into the Repository ..................................................... 23 Integrate the Dual-Boot Script in to the Repository .......................................................... 24 End of script
............................................................................................................................. 24 Create the Post-Installation Script .................................................................................... 25
Prepare for Windows Compute Cluster Server 2003 Deployment ...................................... 27 Prerequisites ..................................................................................................................... 27 Install and Configure RIS .................................................................................................. 28 Add an Image to RIS ........................................................................................................ 28 Inject Drivers ..................................................................................................................... 37 Include Scripts into RIS .................................................................................................... 37
Modify the Answer File ..................................................................................................... 38
Deploy the Operating Systems to the Compute Nodes ........................................................... 39 Deploy SUSE Linux Enterprise Server ................................................................................ 39 Deploy Windows Compute Cluster Server 2003 ................................................................. 40
Switch Operating Systems on the Compute Cluster ............................................................... 49 Boot from Windows Compute Cluster Server 2003 to SUSE Linux Enterprise Server ....... 51 Boot from Linux to Windows Compute Cluster Server ........................................................ 51
Verify the Cluster Setup ........................................................................................................... 53
Windows Compute Cluster Server 2003 ............................................................................. 53 Run a Simple Command .................................................................................................. 53 Run an MPI Command on All Nodes ............................................................................... 55 Log Entries ........................................................................................................................ 59
SUSE Linux Enterprise Server ............................................................................................ 59
Appendix 1 - Background Information ................................................................................. 61 This section provides background material on dual-boot scenarios and WindowsCompute Cluster Server 2003. ......................................................................................... 61 Dual-Boot Scenarios ......................................................................................................... 61 Introduction to Windows Compute Cluster Server 2003 .................................................. 61
Appendix 2 - Deployment Methods for WCCS and Linux ................................................... 63 Windows Compute Cluster Server 2003 Deployment Methods ....................................... 63 SUSE Linux Enterprise Server Deployment ..................................................................... 65
Appendix 3 - Example of WCCS Deployment with ADS ..................................................... 66 Requirements ................................................................................................................... 66
Appendix 4 - Table of Installation Options........................................................................... 70 Note: If you already have SUSE Linux Enterprise Server installed on your HPC cluster, you
can install Windows as a second OS with either RIS or ADS; modifications will be required
to provide partition space for the Windows OS. If you already have Windows Computer
Cluster Server installed on your HPC cluster, you will need to create space for the Linux
partitions. You can then use standard SUSE Linux Enterprise Server installation methods.............................................................................................................................................. 71 Appendix 5 - Related Links ................................................................................................. 72
Links for Windows ............................................................................................................ 72 Microsoft Knowledge Base Articles: ................................................................................. 73 Links for Linux................................................................................................................... 74 AutoYaST: ........................................................................................................................ 74
This document describes the installation and configuration of a high-performance computing(HPC) cluster for a dual boot of Microsoft
®Windows
®Compute Cluster Server 2003 (WCCS)
and SUSE® Linux Enterprise Server 10 SP1. You can use this document as a guide for setting up your own dual boot cluster, although you may need to modify the procedures
described here according to your own particular hardware and software configurations.
Because HPC clusters represent a significant investment of resources, maximizing thebenefits of your existing investment can result in huge savings. By using the same cluster for two operating systems, you gain the flexibility of an additional cluster without having to investin further hardware. There are a number of scenarios in which the ability to switch easilybetween SUSE Linux Enterprise Server and Windows Compute Cluster 2003 on your HPCcluster is an advantage, including proof-of-concept scenarios, application migration scenarios,and test scenarios.
For a dual-boot installation using Windows Compute Cluster Server 2003 and SUSE LinuxEnterprise Server 10 SP1, both must be installed on each of the compute nodes of your cluster. There are several options available for OS installation. In the setup discussed here,Remote Installation Services (RIS) were used for the WCCS installation. RIS, now called theWindows Deployment Services (WDS) legacy mode, is a standard component of WCCS. Itwas installed on the head node of the compute cluster. A dedicated network installation server installed with SUSE Linux Enterprise Server was used to deploy SUSE Linux EnterpriseServer to compute nodes using AutoYaST, the Novell® auto-installation mechanism.
To operate in a dual-boot environment using Windows Compute Cluster Server 2003 andSUSE Linux Enterprise Server 10 SP1, both must be installed on different partitions on thehard disks of all of the compute nodes. For Linux, two hard disk partitions are required. For Windows Compute Cluster Server 2003, one hard disk partition is required.
SUSE Linux Enterprise Server was the first OS deployed to the compute nodes in this setup.It was installed on Partition 1 and Partition 2, leaving the remaining free space, Partition 3, for the Windows Compute Cluster Server 2003 OS. WCCS was then deployed by using RIS. Thisconfiguration was used because it is not possible to create custom partitions with RIS.
Typically, a dual-boot configuration provides a boot menu that lets users choose which OSthey want to boot. A boot menu will not work in a compute cluster dual-boot environment,however, because there are no users at the compute nodes. To switch between operatingsystems, the disk partition of the desired OS is flagged as active and the nodes are rebooted.This process can be automated by executing a task that runs on all of the compute nodes. If the cluster is running Windows Compute Cluster Server 2003, either the compute cluster
administrative GUI or a script can be used to flag the SUSE Linux Enterprise Server partitionsas active. If the cluster is running SUSE Linux Enterprise Server, various methods can beused to run a script on all of the compute nodes to flag the Windows Compute Cluster Server
2003 partition as active. In this setup, the Secure Shell protocol (SSH) was used to run the
script.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
There are other configurations and deployment methods available for a dual-boot HPCcluster, but these were not verified for this document. Manual installation is not describedhere, but it can be easily derived from the provided scripts and installation guidelines. Theinstallation methods and configurations that you choose will depend on the number of compute nodes that you use, and on the order in which you choose to install the operatingsystems. Appendix 2 provides information about installation options including WindowsDeployment Services (WDS); Appendix 3 gives an example of Windows Compute Cluster
Server 2003 deployment with the Automated Deployment Services (ADS).
In this document, it is assumed that all compute nodes and the required infrastructure serversof the compute cluster do not operate with different OS versions concurrently. This istechnically feasible, but it requires a sophisticated infrastructure setup that is beyond thescope of this document. This document does not describe how to dual boot the cluster’s headnode or the infrastructure servers. For these servers, it is possible to use a boot menu tochoose the desired OS at startup. Infrastructure servers may require additional partitions inorder to run the two operating systems.
The focus of this document is the deployment of the Windows Compute Cluster Server 2003
OS and SUSE Linux Enterprise Server 10 SP1 to the compute nodes of a cluster and theoperation of the cluster in a dual-boot environment. Configuration of Linux-based andWindows-based HPC clusters is not described. For information on the installation of an HPCcluster, see the related links in Appendix 5.
This document is intended for Linux administrators and Windows IT professionals who arefamiliar with HPC and who have some familiarity with Windows Compute Cluster Server 2003.For more information about the products and methods described, see the links in Appendix 5.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Before you begin the dual-boot configuration and deployment, be sure to read through all of the procedures and verify that your configuration meets all requirements. You should alsobecome familiar with the product documentation for the operating systems and for thedeployment methods (see Appendix 5 for related links). Be sure to test the deploymentprocedures on one node before deploying the operating systems to the entire cluster. Youshould also verify the operation of your cluster by running a small cluster-aware application.
Note: This document describes one particular installation, in which all infrastructure services,such as DNS, Active Directory
®, and DHCP, were installed on the compute cluster head node.
Your configuration may be different, and these services may be installed on a dedicatedinfrastructure server. You may therefore need to modify the procedures described.
Mixed Compute Cluster Conf igura t ion
A compute cluster uses several types of servers, some required and some optional. Theservers that you need will depend on your cluster configuration.
In this installation, a dedicated deployment server was used for the SUSE Linux EnterpriseServer deployment. RIS, as available on the WCCS head node by default, was used to deploythe Windows Compute Cluster Server 2003 OS. An infrastructure server was not used; allservices were placed on the head node.
Note: If you are not building a cluster for test or evaluation purposes, you may have servicessuch as DNS, Active Directory, and DHCP, on dedicated servers, such as the infrastructureserver described below.
Compute Nodes
• The compute nodes’ BIOS must be configured so that the Preboot ExecutionEnvironment (PXE) is the first boot option.
• If the compute nodes have two or more Network Interface Cards (NICs), PXE mustboot only on the interfaces connected to the compute cluster’s private network.
• Manufacturers’ network drivers may be required for the RIS installation.
Windows Compute Cluster Head Node
One server is required to act as the compute cluster head node. This server can beconfigured to act as a compute node, as well.
• The head node can provide the required infrastructure services, such as DNS, ActiveDirectory, and DHCP; this is an option in a test or evaluation environment, or in asmall cluster running on an isolated network.
Note: Windows Internet Connection Sharing (ICS) can assign IP addresses to thenodes, but you can use a DHCP server on the head node or on the infrastructureserver instead. ICS has an address range limit of 192.168.0, and limits the number of nodes in a cluster to 250.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
• For RIS, the head node must have two NTFS partitions on the hard disk and must beconfigured with at least two NICs.
• Configuration of the infrastructure of the head node is not described in this document.Refer to the Windows WCCS documentation for information on infrastructureconfiguration; for the link, see Appendix 5.
Infrastructure Server
This server is optional in a test or evaluation environment.
• The infrastructure server can provide DNS, DHCP, file services and Active Directoryservices if these services are not provided by the head node due to configuration andperformance concerns.
• The infrastructure server can be configured with two NICs, one for the private networkand one for the public network.
Sof tw are Requirements and Vers ions Used
Windows Compute Cluster Server
This installation used Windows Compute Cluster Server 2003 with Service Pack 1 (SP1),which consists of Windows Server 2003 Compute Cluster Edition (CCE) with Service Pack 2(SP2) and the Microsoft Compute Cluster Pack (CCP) with Service Pack 1 (SP1). The headnode used Windows Server 2003 Standard x64 Edition with Service Pack 2(SP2).
The CCP is a combination of interfaces, utilities, and management infrastructure; it can bepurchased separately and deployed to an existing server running Windows Server. Other supported operating systems for both head node and compute nodes include Windows Server 2003, Standard x64 Edition; Windows Server 2003, Enterprise x64 Edition; Windows Server 2003 R2 Standard x64 Edition; and Windows Server 2003 R2 Enterprise x64 Edition.
SUSE Linux Enterprise Server
For Linux installation and deployment, SUSE Linux Enterprise Server 10 SP1 was used andverified on x64 hardware with serial attached storage (SAS) drives and included the MPICH(MPI) RPMs. The GRand Unified Bootloader (GRUB) was also used.
Hardware Used
This installation used the Microsoft Enterprise Engineering Center’s 64-node Skagit cluster for the verification of this document. There are many possible hardware configurations, and youmay need to modify the procedures based on the equipment that you use.
The Skagit cluster uses:
• 4 HP BladeSystem c7000 enclosures, each with:o At least one GbE2c Ethernet Blade Switch
o At least one 4X DDR IB Switch Module for HP c-Class BladeSystem (IFB)
• 64 HP ProLiant BL465c G1 Server Blade nodes, each with:
o 2 dual-core AMD Opteron Processors 2216 HE at 2.40 GHz (1 MB L2)
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
2 embedded NC370i Multifunction Gigabit Server Adapters
Prepare for Window s Comput e Cluster Server 2003 and SUSE Linux
Enterpr ise Server 10 SP1 Dual Boot
Before you begin the deployment and configuration of the compute cluster, verify thefollowing:
Check whencompleted
Prerequisites
The PXE boot option is disabled on all of the network adapters on the head node.
If using a Windows infrastructure server, the PXE boot option on all NICs on theinfrastructure server must be disabled.
The PXE boot option must be disabled on the Linux deployment server.
All drivers that are not integrated into the Windows setup files by default must beavailable.
All of the compute cluster hardware must be set up and working.
A naming schema must be provided.
An IP address schema must be provided.
The DHCP service must be installed on the compute cluster head node or on theinfrastructure server.
The DHCP scopes for the cluster’s private and MPI networks must be created.
The DHCP scope options (at a minimum, DNS servers and DNS domain name)must be created.
The DHCP server must be authorized in Active Directory.
The MAC addresses of the network adapters connected to the private network on allnodes must be available. These are required for configuring the Linux deploymentserver and the Windows DHCP service.
The DHCP service must be configured to use static IP addresses for nodeidentification.
DNS must be installed and configured on the head node or on the infrastructureserver.
Active Directory must be installed on the head node or on the infrastructure server.
The DHCP service must be configured so that it serves the nodes on both theprivate network and the MPI network.
All required hotfixes must be available.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Begin dual-boot configuration by preparing for the OS deployment to the compute nodes. Besure to read through this entire document before beginning installation. It is also important tobecome familiar with the product documentation; for related links, see Appendix 5.
Prepare for SUSE Linux Enter pr ise Server Deployment
In this setup, SUSE Linux Enterprise Server 10 SP1 was deployed as the first OS from adedicated deployment server by using AutoYaST, the Novell automated installationmechanism.
Note: Because AutoYaST uses unicast-based and file-based installation, it might not scalesuccessfully for a very large number of nodes (the maximum simultaneous installations isusually 40-50 nodes).
To prepare for SUSE Linux Enterprise Server deployment, configure the deployment server,create the AutoYaST control file, and integrate the Linux HPC software into the repository.
Configure the Deployment Server
The deployment server must be connected to the private network of the cluster, and must beassigned a static IP address that is in the private network range. The following sectiondescribes the procedures that are used to configure the deployment mechanism on SUSELinux Enterprise Server 10 SP1. It is possible to use other hardware and softwareconfigurations. You should modify the procedures based on your particular configuration.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
1. Install SUSE Linux Enterprise Server 10 SP1 on the deployment server.
a. Disable the the firewall after the installation. To do this, launch YaST (Computer->YaST) and open Security and Users->Firewall. In the ServiceStart section choose Manually, then choose Stop Firewall Now in theSwitch On and Off section. Click Next and then Accept to save settings andshut off the firewall. Once you've got the configuration set and have
successfully performed a PXE boot installation, you can re-enable the firewallwith the allowed service ports, if desired (they are: NFS 2049, TFTP 69,DHCP Server 547 and others as needed).
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
2. Configure a network installation source using YaST Installation Server module
a. Launch YaST (Computer->YaST), browse to the Miscellaneous section andopen Installation Server. The first time you launch the Installation Server module, you'll be prompted to set up an HTTP, FTP or NFS installationsource. Choose NFS. In this example, an “/install” directory was created atthe root of the file system to contain the installation sources.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
b. On the following screen you are prompted to enter a Host Wild Card andOptions for the NFS share. Enter “*” (asterisk) in the Host Wild Card field,“ro” in the Options field and click Next. (If the “Unable to write idmapd.conf”error pops up you can safely ignore it)
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
c. After configuring the installation server, you are now ready to add installationsources. Click Add to add a new installation source. Enter a Source Name (“SLES10SP1-64” was used for this example) and click Next. Choose asource to import the installation source. In this example, the DVDROMwas used (you can use the same media that was used to install the OS onthe server), but a CDROM or the distribution ISOs could be used just as well.On clicking Next, you'll be prompted to insert the first CD (or DVD). Insert the
first DVD of SUSE Linux Enterprise Server 10 SP1 and click Continue (Thecopy process can take a while).
After the first DVD or DVD ISO image has been copied, you'll be prompted toenter the second DVD (or DVD ISO). This step can be safely skipped(choose Skip) as the second DVD only contains the source RPMs. You can,however, copy the source RPMs as well if desired. In the case of CDs or CDISOs, you'll want to copy at least the first four. You should now be back at theInstallation Server main screen. Click Finish here to wrap up theconfiguration.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
d. You have now created an NFS installation server with a SUSE LinuxEnterprise Server 10 SP1 installation source. You can verify that the sourcewas created correctly by mounting the NFS share from another Linux clienton the network. Type the following commands (replace 192.168.0.200 with
the address of your installation server):
mount 192.168.0.200:/install /mnt
ls /mnt
End of script
You should see the contents of the installation server under the /mnt directory atthis point. Type the following commands to unmount the share:
umount /mnt
End of script
3. Configure a PXE (Preboot Execution Environment) server
Note: If you are unable to place a DHCP server on the same subnet as your HPC
nodes (maybe there is an existing DHCP infrastructure), consider using custom bootCDs to kickoff network installations. Skip to step 6 for details on how to create acustom boot CD.
a. To set up a PXE server (on the same installation server) you'll need to createa PXE directory structure and some accompanying configuration files. Youwill also need to enable and configure DHCP and TFTP services. To start,launch YaST, browse to Software and open Software Management. In theSearch field, type “syslinux” and click Search. If the package is not yetinstalled, check the box. Follow the same steps to install “dhcp-server”,“tftp” and “xinetd”. Click Accept to install all of the packages. Once theinstallation is complete, issue the following commands in a terminal:
In the above commands, you first created a two-level directory structure,/tftpboot/pxelinux.cfg. Second, you gave “write” permissions to all users in thepxelinux.cfg directory. Next, you copied “pxelinux.0” (the network bootstrapprogram) to the /tftpboot directory. Finally, you copied the Linux kernel and
the initrd file (initial ramdisk) from the installation source to the /tftpbootdirectory.
b. In the /tftpboot/pxelinux.cfg directory, create two files, “default” and“boot.msg”. Following are samples of the two files:
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
4. Configure TFTP (Trivial File Transfer Protocol)
a. To set up TFTP, launch YaST, browse to Network Services and open TFTPServer. If you are prompted to install the tftp RPM, click Continue. At theTFTP Server Configuration, select Enable, enter “/tftpboot” as the BootImage Directory and click Finish. The TFTP server is now configured andenabled.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
b. Edit /etc/xinetd.d/tftp. This step allows files to bewritten to the tftp server. You can use this
functionality to help prevent endless installation loops
later.
# default: off
# description: tftp service is provided primarily for booting or when
a \
# router need an upgrade. Most sites run this only on machines acting
as
# "boot servers".
service tftp
{
socket_type = dgram
protocol = udp
wait = yes
user = root
server = /usr/sbin/in.tftpd
server_args = -s /tftpboot -c -u root
disable = no
}
End of script
5. Next, configure the DHCP server by editing the file /etc/dhcpd.conf. (If a DHCP
server is already in place on your network, you may be able to leverage it using thesame or similar parameters to direct PXE requests to your TFTP server). A samplefile is shown below; modify the highlighted parameters based on your particular configuration, as follows.
α. Adjust the subnet, netmask, and range as required.
β. Adjust the next-server option so that it points to the IP address of the SUSELinux Enterprise Server deployment server.
χ. Add an IP reservation for each node on your cluster by adding a host name
Note: Follow this step only if you are unable to create a PXE Boot environment onyour HPC subnet or private network
a. If you don't have control over the DHCP server on your subnet or if you're notable to setup a TFP server for one reason or another, all is not lost. You cancreate a custom boot CD with hard-coded install and autoyast parameters. To do
this, simply download the CD1 ISO of SUSE Linux Enterprise Server 10 SP1,available here: http://www.novell.com/linux/download_linux.html
b. Loopback mount the ISO and copy the contents to a local directory, commandsbelow (as root):
mount -o loop SLES-10-SP1-x86_64-CD1.iso /mnt
mkdir /tmp/cdroot
cp -a /mnt/* /tmp/cdroot
End of script
c. Edit the isolinux.cfg, found in boot/x86_64/loader from the root of “cdroot”, toinclude the install and autoyast parameters shown in the file snippet below:
Next, install a single node leveraging your Network Installation/PXE environment.
1. Select Network Boot at boot time or make sure that the Network Boot is set to 1st
inthe BIOS boot order.
a. Type “2” and <enter> at the “boot:” prompt to perform a manual, network-based
installation. Go through the install, selecting the options appropriate for your environment. This system will become the template for others that you will deploy.(You may need to escape <ESC> PXE boot after the first installation phase. Thesystem will reboot after the first installation. Escaping the PXE boot sequence willprevent the system from starting the installation. More on how to automate thislater in the document.)
b. At the end of the installation, check the Clone This System for Autoyast checkbox. This will generate an AutoYaST file based on the installationinformation that was just used. The generated file, called “autoinst.xml” can befound in “/root” after install.
Additional configuration may be needed after install to correctly setup NFS mounts, user authentication and install additional RPM packages. You can re-export these changes to anAutoYaST file. To do this launch YaST (Computer-YaST), browse to Miscellaneous andselect Autoinstallation . You can also export the current configuration of the system bychoosing Tools->Create Reference Profile. The Autoinstallation module also allows you toedit existing AutoYaST files without knowing the underlying xml syntax.To edit existingAutoYaST file, choose File->Open. When finished editing choose File->Save.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
You can also edit the AutoYaST file directly using a text editor. A sample AutoYaST file isprovided in Appendix 8 for reference. Once you have customized the AutoYaST file for your environment, save the file to /install/hpcautoyast.xmlon the Network Installation
Server. The relevant sections from the sample AutoYaST file are shown below with guidancefor customizing the script.
1. Manually add the <pxe> section to the saved control file. This will cause AutoYaST to
upload a file to the TFTP server; the name of the file is the IP address of the node.This procedure prevents a PXE boot loop.
2. Replace 192.168.0.200 with the IP address of your Linux deployment server. The
relevant portion of the sample script is shown below, with the IP address that must bereplaced highlighted.
3. Modify the portion of the control file that will partition the hard disks of the computenodes to reflect your desired partition sizes and the device name of your deploymentserver.
Note: The SUSE Linux Enterprise Server OS will be located on Partition 1 of thedisks of the compute nodes; Swap will be installed on Partition 2. In this installation,Partition 1 was configured to be 10 GB and Partition 2 was configured to be 2 GB.
You can adjust these sizes to suit your configuration. The disk space remaining after Linux installation will be left free for the Windows Compute Cluster Server 2003installation.
4. Next, use a custom prescript, killMBR.sh, which will flush the partition table
when AutoYaST is started. The relevant portion of the control file is shown below. Thehighlighted IP address should be changed to reflect the address of your deploymentserver.
Note: This step was necessary in this setup to ensure that the SUSE Linux Enterprise
Server installation worked with the cluster’s RAID controller. It may not be necessaryto use a custom prescript with your configuration.
1. Integrate the root´s public key into the repository to ease SSH access later (accept the
defaults and leave the password empty on generating the key).
mkdir –p /install/custom/ssh
ssh–keygen –b 1024 –t dsa
cp /root/.ssh/id_dsa.pub /install/custom/ssh/
End of script
Integrate the Dual-Boot Script in to the Repository
1. Create the directory by running the command:
mkdir –p /install/custom/hpcscripts
End of script
The directory is used for the script that switches the OS from Linux to Windows. Itmust be included in the repository so that it is automatically deployed to all computenodes.
2. Create the following input file for fdisk, and save it as
/repository/custom/hpcscript/MBRWindows-3.txt.
The script toggles the bootable flag of Partition 2 from bootable to not-bootable (a 1).
It then toggles the bootable flag of Partition 3 from not-bootable to bootable (a 3),
and writes (w) the changes to the partition table.
a
1
a
3
W
End of script
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Prepare for Window s Comput e Clust er Server 2003 Deployment
A version of the Remote Installation Service (RIS), now called the Windows DeploymentServices (WDS) legacy mode with the application of SP1 to CCP, is integrated into the CCPand is installed on the compute cluster head node by default. It does not require additionalhardware.
Note: Do not be concerned when you see Windows Deployment Server on the ComputeCluster Administrator page. Starting with CCP SP1, RIS is referred to as WDS legacy mode.Thus, while the wizard GUI now reads ‘Windows Deployment Services” (as shown in thescreen shots later in this document), the process that the wizard uses and the steps describedin this document are nonetheless based on RIS. In this document, the deployment method isreferred to as RIS, and the procedures will work as you expect them to under RIS.
Other installation methods for Windows are available. For more information, see links in Appendix 5.Prerequisites
Before you begin a deployment of Windows by using RIS, verify the following.
Check whencompleted
Prerequisites
Every item on the checklist for Windows Compute Cluster Server 2003deployment with RIS, found in Appendix 6, must be completed.
The network adapters of the compute nodes must support PXE.
PXE must be selected as the first boot device in the hardware’s BIOS for theprivate network.
PXE must be disabled on the compute node NICs that are connected to theMPI network.
The compute cluster head node must be installed and running.
A Windows Compute Cluster Server 2003 infrastructure (Active Directory,DNS, and DHCP) must be available to the head node and to the computenodes.
A dedicated NTFS partition must be available for the RIS setup file.
The CCP must be installed and configured on the head node.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Begin by installing RIS on the head node of the compute cluster. Refer to the Windows Server 2003 Compute Cluster Edition Product Documentation for a standard RIS setup. This can befound at:
1. On the To Do List page of the Compute Cluster Administrator, click Install RIS(wizard) on the Remote Installation Services (RIS) tile.
2. Note the tasks listed on the Before You Begin page.
3. On the Configure RIS page, click Install Remote Installation Services.
4. On the Update summary page, verify that the message Install remote installationservices is displayed.
5. Observe the progress of the RIS installation on the Configuring Components page.When the installation is successfully completed, the Update Complete page willappear.
6. To complete the installation, click Finish.
Add an Image to RIS
Next, add an image to RIS. This is the image that will be deployed to the compute nodes.
Note: When included, screen shots immediately follow the step that they illustrate. They can
be used for verification of the procedure described.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
5. Adjust the information in the Image Name, Description, and Architecture fields, asneeded.The files that are required for the RIS image will be copied from the WindowsCCE CD-ROM.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
11. On the View Summary page, click Finish. Verify the results, and then click Close.
Inject Drivers
If required, include additional hardware drivers into the RIS setup. For more information, seeMicrosoft Knowledge Base (KB) article 246184, “How to Add Third-Party OEM NetworkAdapters to RIS Installations.”
Include Scripts into RIS
Next, modify the default RIS installation to support the dual-boot configuration. A script isrequired to change which partition is active on the compute nodes. Include this script in theRIS installation tree on the head node; the script will then be distributed to the compute nodes
during the deployment procedure. The script is used by Windows to mark Partition 1—theLinux partition that has Grub installed on its boot sector—as the active partition.
1. On the head node, open Windows Explorer and change the current directory toE:\RemoteInstall\Setup\English\Images\<your-image-
name>\$OEM$\$1.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
2. Within this directory, create a new directory named HPCSCRIPTS.Create the text
file, mbrlinux-1.txt, as shown below. This file will be used as input for
diskpart.exe; it contains the script that is used for switching operating
systems.
SELECT DISK 0
SELECT PARTITION 1
ACTIVE
End of script
Modify the Answer File
Next, modify the RIS setup answer file. For a sample answer file, see Appendix 8. Save abackup copy first, and then edit and save the file in Notepad.
1. Make a backup copy of the file, namedE:\RemoteInstall\Setup\English\Images\<your-image-name>\<your-
architecture>\templates\Ristndrd.sif .
2. Open the file with notepad.exe.
3. Modify entries in section [RemoteInstall] as follows.
Repartition=Yes
UseWholeDisk=No
End of script
4. Modify entries in section [data] as follows.
DisableAdminAccountOnDomainJoin=0
End of script
5. Provide the Password for local Administrator in the section [GuiUnattended].
This enables you to log on to a compute node locally if you need to troubleshoot.
AdminPassword=”mypassword”
End of script
Modify the path to the CCP setup files. When Windows Compute Cluster Server 2003is installed on an empty Partition 3, drive E is usually used. The drive may be differentin your configuration; modify the path accordingly.
Deploy the Operat ing Sys tems t o t he Comput e Nodes
After the deployment server is configured for Linux and RIS is installed on the head node, theoperating systems can be deployed to the compute nodes of the cluster.
Note: The Linux OS must be deployed to Partition 1 and Partition 2 of each compute node asthe first OS. This will leave all of the remaining free space for the Windows Compute Cluster Server 2003 installation. This configuration is necessary because RIS, as installed on thehead node, does not let you to create a custom partitioning scheme.
It is important to test your configuration by deploying to one node or to a small number of nodes before deploying to the entire cluster.
It is possible to use other configurations and deployment methods, but these are morecomplex and will require modifications to the procedures described in this document. For more information about alternate deployment methods, see Appendix 2.
Deploy SUSE Linux Enterpr ise Server
Begin the deployment of the operating systems to the compute nodes of the cluster bydeploying SUSE Linux Enterprise Server as the first OS from the Linux deployment server toPartition 1 (the root partition) and Partition 2 (the swap partition) of each compute node.
Note: This document focuses on the installation of the basic Linux OS to the compute nodes.It does not describe how to configure a Linux cluster or how to set up a Linux head node.
1. Start the SUSE Linux Enterprise Server deployment.
2. Ensure that the RIS service has been stopped on the WCCS head node.
3. Stop the Windows DHCP services on the WCCS head node (or on the infrastructurenode, if used).
4. Start the dhcpd daemon on the SUSE Linux Enterprise Server deployment withthe command rcdhcpd start.
5. Boot all compute nodes into PXE boot.
Note: Depending on the number of nodes you have, you might choose not to start theinstallation simultaneously on all of the compute nodes (usually 40 simultaneous
installs in the upper limit).
6. Verify that the Linux installation has started.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
7. After all nodes have been deployed, shut down the dhcpd daemon by using thecommand rcdhcpd stop. SUSE Linux Enterprise Server should now be installed on all
of the compute nodes.
Note: To verify that all compute nodes are running Linux, you can create a shell scriptthat uses SSH and attempts to log on to all of the nodes sequentially.
8. If it is necessary to restage a Linux computer, delete the corresponding file
/tftpboot/pxelinux.cfg/AABBCCDD , where AABBCCDD is the hexadecimal notation
for the node´s IP address. Refer to the AutoYaST documentation for further information;see the link in Appendix 5.
Note: To verify that all compute nodes are running Linux, you can create a shell script
that uses SSH and attempts to log on to all of the nodes sequentially.
Deploy Window s Comput e Clust er Server 2003
When Linux is installed on the compute nodes, deploy Windows on Partition 3 with RIS. Thedrive on which Windows Compute Cluster Server 2003 is installed will usually not appear as
drive C. Typically there is a CD drive in addition to the two Linux partitions, so Windows willbe installed on drive E. This is usually transparent to the Windows Compute Cluster Server 2003 OS and to the applications, but it must be considered for hard-coded paths in setuproutines and during the runtime of applications and scripts.
Note: This document focuses on the installation of the basic Windows Computer Cluster Server to the compute nodes. It does not describe how to configure a Windows cluster or howto set up a Windows head node.
Prerequisites
Before you begin the Windows Compute Cluster Server 2003 deployment, verify the following.
Check when
completed
Prerequisites
SUSE Linux Enterprise Server must be installed on the compute nodes.
Grub must be installed on the boot sector of the Linux partition, not onthe MBR.In this scenario, the Linux AutoYaST control file installs Grub on the bootsector of the Linux partition.
The Linux boot partition must be active. If the Linux boot partition is notactive, the system may be inaccessible.In this scenario, the Linux AutoYaST control file flags the Linux partitionsas active.
The Linux deployment services must be stopped. Use the commandrcdhcpd stop.
The Windows-based DHCP service on the head node or on theinfrastructure server must be started.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
7. Reboot the compute nodes. Because all of the nodes are running Linux, you can useSSH to initiate a reboot remotely. For an example, see Appendix 9. You can
simultaneously boot any number of nodes.
8. Connect to the console (by using the chassis’s hardware) of several of the computenodes to verify that they boot into PXE and that the Windows Compute Cluster Server 2003 installation procedures have started. Continue when the installation has finishedon all of the compute nodes.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
2. Select all of the nodes, right-click, and then select Approve.
3. Check that all of the nodes are selected, right-click, and then select Resume.
Note: If you must reinstall Windows Compute Cluster Server 2003, first delete Partition 3 onall of the compute nodes, activate Partition 1, and then delete the selected compute nodemachine objects in Active Directory. If you do not delete Partition 3, the next time thatWindows is installed, the Windows partition will be assigned the drive letter C.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Sw i tc h Operat ing Sys tem s on the Comput e Clus ter
If a computer’s hard disk has no boot loader in the MBR and has two operating systems on itspartitions, the computer’s BIOS will execute the boot loader that is in the active partition andboot the OS that is installed on the active partition. The compute nodes can be forced to bootthe desired OS by using a command or a script to flag the desired OS partition as active. After flagging the partition as active, all compute nodes must be rebooted to start the alternate OS.
To switch between Windows Compute Cluster Server 2003 and SUSE Linux EnterpriseServer on the compute nodes, remotely change which hard disk partition is active on all of thecompute nodes, and then reboot. You can change the state of the partitions with the scriptsthat were deployed to the compute nodes during OS installation, or you can change the stateof the partitions manually.
If Windows Compute Cluster Server 2003 is the active OS on the nodes, use the ComputeCluster Administrator GUI to remotely execute the diskpart script that was integrated
during the RIS installation, and then reboot the compute nodes.
If SUSE Linux Enterprise Server is the active OS, use your favorite method to remotely
execute the fdisk script that was integrated during the Linux installation, and then reboot the
compute nodes. In this scenario, SSH was used to remotely execute the fdisk script on the
compute nodes. The fdisk script used the answer file script /hpcscripts/MBRWindows-
3.txt to change the active partition. If you have installed a dual boot setup for the
infrastructure node or the head node, you can manually set the active OS. For sample scripts,see Appendix 7.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Boot f r om Window s Comput e Cluster Server 2003 to SUSE Linux
Enterpr ise Server
If Windows Compute Cluster Server 2003 is the active OS, a Windows diskpart.exe script
must be run on all compute nodes. The diskpart.exe script will use the text file
mbrlinux-1.txt as input; this text file is available on the compute nodes when they are
running Windows Compute Cluster Server 2003 because it was deployed by RIS along withthe OS. The script marks the Linux partition as active; it must be used because Grub isinstalled on the boot loader of Partition 1.
1. Using the run command in the Compute Cluster Administrator GUI or using the
WCCS clusrun command, run the following script on all compute nodes by using
the following syntax.
Diskpart.exe /s E:\HPCSCRIPTS\mbrlinux-1.txt
End of script
2. Reboot all of the compute nodes from the Windows Compute Cluster AdministrationGUI by selecting all of the nodes and then running the following command. Theparameter xx is the timeout period to shutdown. The default is timeout period is 30
seconds. The reboot can also be scripted with the WCCS clusrun command.Shutdown.exe /r /f /t xx /c ”Shutdown to boot into Linux”
End of script
Boot f rom L inux t o Window s Compute Cluster Server
If SUSE Linux Enterprise Server is the active OS, change the active partition remotely byrunning the script fdisk with the input file MBRWindows-3.txt. This script is available on
the compute nodes when they are running SUSE Linux Enterprise Server because it isdeployed along with the SUSE Linux Enterprise Server by the deployment server. You mustmodify the device name in the script according to your particular hardware configuration.
1. Run the command ‘fdisk /dev/cciss/c0d0 < /hpcscripts/MBRWindows-
3.txt’ on all of the compute nodes.
Alternately, you can run a script that remotely executes the changes to the partitiontable.
After Windows Compute Cluster Server 2003 has been deployed to all of your computenodes, verify that the cluster is functional by running a simple command and an MPI
command on all nodes.
Note: When included, screen shots immediately follow the step that they illustrate. They canbe used for verification of the procedure described.
Run a Simple Command
1. Navigate to Node Management in the Compute Cluster Administrator.
2. Select all of the nodes, right-click, and then select Run Command.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Next, execute an MPI application that tests the cluster setup. Schedule and execute a parallel job on the entire cluster to verify functionality. The job is a simple parallel application thatapproximates the value of π (pi) based on numerical integration.
Note: When included, screen shots immediately follow the step that they illustrate. They canbe used for verification of the procedure described.
1. Download the file BATCHPI.ZIP titled “Batchpi – a simple MPI application, for testing
purposes” from the Windows HPC Code gallery:http://windowshpc.net/files/18/test_apps/entry13.aspx
2. Create the directory C:\testsmpd on the head node.
3. Create a file share to the C:\testsmpd directory by using the name testsmpd.
To create a file share:
a. Open Windows Explorer, and then locate the c:\testsmpd folder to which you
will add a new share name.
b. Right-click c:\testsmpd folder, and then click Sharing and Security.
c. Click Share this folder.
d. Click Permissions.
e. Set the permissions to Full Control for Everyone.
4. Unzip the contents of the file BATCHPI.ZIP into c:\testsmpd.
5. From the Start menu, start the Cluster Job Submission and Monitoring Console.
6. Create a new job submission by clicking on File, and then clicking on Submit Job.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
11. You can save the job as a template by clicking the Save As Template… button onthe Tasks tab, at the bottom of the dialog box. Submit the job by clicking the Submit button at the bottom of the Task properties dialog box.
12. After you submit the job, you will see the job queue in the Cluster Job Submissionand Monitoring Console. Note that the initial job status will be Running. After a fewminutes, the status will change to Finished or Failed.
13. If the job failed, make sure that you have entered all data correctly, and then tryagain. If the problem persists and you have had failures with other indicators of cluster health, examine the log entries for clues to the problems.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
15. When you have finished, remove the file share that you created to avoid securityissues.
Log Entries
View the Compute Cluster warnings and errors; this will assist you in solving real or potentialissues with your cluster.
1. Open the Event Viewer and look for warnings or errors in the Application andSystem logs.
2. If any warning or error log entries are present, double-click on the error message for additional details and guidance.
SUSE Linux Enter pr ise Server
The configuration of SUSE Linux Enterprise Server based compute cluster is not discussed inthis document. To verify that your Linux compute cluster is running, boot all of the computenodes into Linux by using the provided scripts, and then run a sample MPI command.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
This document describes the installation and configuration of an HPC cluster for a dual boot
of Windows Compute Cluster Server 2003 and SUSE Linux Enterprise Server. It shows the
deployment of the operating systems to the compute nodes, the procedure for switching
between the Windows Compute Cluster Server 2003 and SUSE Linux Enterprise Server on
the compute nodes, and the verification of the cluster setup. Every IT environment is different,and this guide can serve as a basis for setting up your own dual-boot compute cluster. If you
need additional guidance, see the related links section in Appendix 5 for resources.
This installation used SUSE Linux Enterprise Server, and AutoYaST as the Linux deployment
method. Other deployment mechanisms exist including block level imaging and multicast
imaging, available in Zenworks® Linux Management,that might be better suited to your
environment.
In this installation, RIS, also called WDS legacy mode, was used as the deployment method
for the Windows OS. Other Windows Compute Cluster Server 2003 deployment methods are
available, and some might be more appropriate for your particular needs. RIS is best for smallto medium-sized clusters. For larger clusters, ADS might be a better option because it uses
the multicast streaming technology. WDS, the next generation of RIS, might also be a good
option for deploying Windows Compute Cluster Server 2003 to your (small to medium-sized)
cluster. See Appendix 2 for a discussion of Windows deployment methods.
The installation and configuration of a HPC compute cluster involves many diverse elements.
This document does not attempt to cover all of the procedures, but focuses only on the dual-
boot configuration and operation. For more information about topics not covered in this
document, refer to the links in Appendix 5.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
This section provides background material on dual-boot scenarios and Windows ComputeCluster Server 2003.
Dual-Boot Scenarios
High-performance computing (HPC) is used to solve computational problems that requiresignificant processing power and quick access to very large amounts of data. HPC clustersmake it possible for businesses and researchers to solve complex computational problemsfaster than ever before. HPC is used extensively in areas such as financial modeling,computational fluid dynamics, and biomedical research. Over the past ten years, industry-standard hardware and software have dramatically improved in price, performance, andusability. HPC is now affordable and available to a broad range of organizations andbusinesses, and a Windows-based platform is a viable and attractive choice.
HPC clusters can represent a significant investment of your resources. Your HPC cluster canbe composed of tens, hundreds, or even thousands of computers. Maximizing the benefits of your current investment can bring tremendous savings; by using the same cluster for twooperating systems, you gain the flexibility of another cluster without the hardware outlay.There are a number of scenarios in which the ability to easily switch between SUSE LinuxEnterprise Server and Windows Compute Cluster (WCCS) on your HPC cluster is a major advantage.
Moreover, if you have Linux-dedicated hardware and you want to develop WCCS skills, adual-boot cluster can be very helpful. With a dual-boot cluster, you can try Windows ComputeCluster and become familiar with the user interfaces and commands without investing in anew set of hardware. A dual-boot environment also enables you to test usage scenarioswithout fully disengaging your existing operating system. You can observe and fine-tune your application on Windows Compute Cluster while maintaining your existing environment.
Development organizations that build applications that run on both the Linux operating systemand on Windows Compute Cluster Server 2003 can benefit from a dual-boot environmentbecause it lets you test applications on both operating systems without investing in two sets of hardware. The ability to run applications on Linux and Windows Compute Cluster Server 2003on a single set of hardware can give you additional flexibility without further investment.
Introduction to Windows Compute Cluster Server 2003
HPC clusters are computing systems comprised of multiple processors that are linked
together into a single system by using commercially available interconnects and commodity
software. Because of their flexibility, their computing power, and their relatively low cost, HPC
systems are increasingly dominating the world of supercomputing.
Windows Compute Cluster Server 2003 (WCCS) is the Microsoft solution for HPC clusters.
WCCS brings together all of the needed HPC components in a single package that was
designed and developed to run together. WCCS is fully supported and tested, and it can
leverage any existing Windows infrastructure. WCCS offers ease of management and
enhanced security; it also provides features for interoperability with Linux operating systems.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Windows Compute Cluster Server 2003 with Service Pack 1 (SP1) contains two CDs:
• The first CD contains Windows Server 2003 Compute Cluster Edition (CCE) with
Service Pack 2 (SP2). The CCE can be purchased separately to be used as a
dedicated HPC operating system.
• The second CD contains the Microsoft Compute Cluster Pack (CCP) with ServicePack 1 (SP1), a combination of interfaces, utilities, and management infrastructure.
The CCP can be purchased separately and deployed to a server computer running
Windows Server. Supported operating systems besides Windows Server 2003
Compute Cluster Edition for both head node and compute nodes include:
o Windows Server 2003, Standard x64 Edition.
o Windows Server 2003, Enterprise x64 Edition.
o Windows Server 2003 R2 Standard x64 Edition.
o Windows Server 2003 R2 Enterprise x64 Edition.
The remote administration and job scheduling components are installed automatically
on the head node of the compute cluster but can also be installed on a remote
workstation to simplify administration and scheduling. The supported operatingsystems for installation of the remote components are:
o Windows Server 2003 with Service Pack 1 or 2 (SP1 or SP2) or R2
o Windows XP Professional with Service Pack 2 (SP2)
o Windows XP Professional x64 Edition
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Appendix 2 - Deployment Met hods for WCCS and Linux
To set up an HPC cluster, you must deploy the OS to all of the compute nodes. For a dual-
boot configuration, you must deploy two operating systems to each compute node. There are
many options available for deploying Windows Compute Cluster Server 2003 and Linux to
your HPC compute cluster nodes. This installation used RIS to deploy Windows Compute
Cluster Server 2003 and AutoYaST to deploy SUSE Linux Enterprise Server. AutoYaST
enables unattended and automated installation by using PXE; it is based on YaST2, an OS
setup and configuration tool that is featured in SuSE Linux Enterprise distributions.
This section of the appendix discusses some of the options available for Windows Compute
Cluster Server 2003 and SUSE Linux Enterprise Server deployment. Your configuration might
require you to use different methods from those used in this installation.
Windows Compute Cluster Server 2003 Deployment Methods
There are several methods available for the deployment of the Windows Compute Cluster
Server 2003.
Remote Installation Services
Remote Installation Services (RIS) is a change-and-configuration-management feature that
supports remote installations of Windows Server 2003 operating systems. You can use RIS to
provide on-demand, image-based OS installations. This occurs over a network connection
from a RIS server to a client compute, which means that you can initiate RIS-based
installations remotely on a computer without being physically present at the computer. RIS
eliminates the need to use either the operating system CD or a floppy disk containing a
Winnt.sif file. You can use RIS to remotely set up new Windows Compute Cluster Server
2003-based computers by using a RIS network-shared folder as the source of the Windows
Compute Cluster Server 2003 files. You can install the OS on remote boot-enabled client
computers. Client computers are connected to the network, and are then started by using a
PXE-capable network adapter or remote boot disk. RIS can simplify the management of
operating systems and applications and can improve failure recovery.
RIS requires several supporting technologies that identify computers on the network. The
following three technologies are included as components of Windows:
• DHCP
• DNS
• Active Directory
DHCP and DNS do not need to be Microsoft versions in order to work with RIS. You must use
the Microsoft version of Active Directory when using RIS, however. DHCP, DNS, and ActiveDirectory can be installed (along with RIS) on one server or on several servers. When you
promote a server to a domain controller, Active Directory is automatically installed. RIS also
depends on a unique identifier that computer manufacturers assign to each computer. This
identifier is called the globally unique identifier (GUID) or the universally unique identifier
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Automated Deployment Services (ADS) combines the speed of traditional imaging with the
flexibility of script-based installation, and is an add-on to Windows Server 2003 Enterprise
Edition. It provides a solution for rapidly deploying the Windows OS onto bare-metal serversacross large, scaled-out installations. With support for script-based mass server
administration, ADS also enables you to administer hundreds of servers as if they were one.
If the number of compute nodes is large, ADS is the preferred method for Windows Compute
Cluster Server 2003 image deployment. ADS uses multicast as transport, and therefore can
accelerate the image deployment. Multicast, a true broadcast, relies on multicast-enabled
routers to forward the packets to all client subnets. However, ADS is not integrated into the
automatic Windows Compute Cluster Server 2003 setup.
First, build the Windows master installation image that ADS will use for deployment of the
compute nodes.
1. Install the Windows Compute Cluster Server 2003 OS on the reference compute
node hardware. This is the master installation image. You can use the standard
installation method. Keep the Windows Compute Cluster Server 2003 partition small;
it can be enlarged later during installation, if necessary.
2. Install all required drivers and update Windows Compute Cluster Server 2003 with
latest service pack and security fixes.
3. Include Sysprep files as described in the ADS documentation.
4. Install the Window Computer Cluster Server 2003 and the Microsoft .NET Framework
version 2.0.
5. Include the script for dual boot, mbrlinux-1.txt, to the directory
c:\hpcscripts. This script is provided in Appendix 7.
6. Install any additional software that you would like deployed with the OS.
7. Add the computer to the domain.8. Boot the computer into the ADS Deployment Agent to start the image capture. It
might be necessary to inject drivers for storage or NICs into the ADS Deployment
Agent. Refer to the section titled Deployment Agent Builder Service in the ADS
Documentation. For more information, see the links in Appendix 5.
Note: For more flexibility in the setup process, use Windows PE instead of the standard ADS
deployment.
Create and Customize ADS Task Sequence for Image Capture
Consult the ADS product documentation for complete instructions for image capture. After thesequence for image capture is created, change the corresponding .xml file.
By default, the ADS imaging process tries to locate the OS on Partition 1 of the hard disk.
Windows is installed on Partition 3, so you must change the path in the corresponding .xml
sequence file.
1. Open the image-capture sequence .xml file with an editor and locate the following
Next, follow the instructions in the ADS product documentation to deploy the WindowsCompute Cluster Server 2003 images to the compute nodes. Reboot the cluster nodes andapprove them in the Windows Compute Cluster Administration GUI or with a WCCS clusrun
script.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Note: If you already have SUSE Linux Enterprise Server installed on your HPCcluster, you can install Windows as a second OS with either RIS or ADS;modifications will be required to provide partition space for the Windows OS. If youalready have Windows Computer Cluster Server installed on your HPC cluster, youwill need to create space for the Linux partitions. You can then use standard SUSELinux Enterprise Server installation methods.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
Appendix 6 - Checkl ist for WCCS Deployment w it h RIS
Use the following checklist to prepare for the deployment of WCCS. After completing theseprocedures, you can begin the additional configuration necessary to prepare for a dual-bootenvironment.
Check when
completedProcedure
Review hardware and software requirements for WCCS.
Determine the topology that you will use for your compute cluster.
Configure the head node:
Install the supported operating system.
Create a second disk volume to store the installation images.
Join the head node to an existing Active Directory domain; if there
is no Active Directory domain available, create a new domain.
Install MMC 3.0 and any required hotfixes. Consult the Microsoft
Knowledge Base for updates about required hotfixes for WCCS.
Install the Compute Cluster Pack on the head node.
Configure the network topology on the head node.
Enable or disable the Windows Firewall on the public network
interfaces of the cluster.
If using RIS as the automated deployment method for your
compute nodes, enable it from the Compute Cluster Pack. Add
one or more installation images, and configure the images with
valid product keys and identifying descriptions.
Add the compute nodes to your cluster, by using either the
automated method or the manual addition method.
Approve the compute nodes.
Define the cluster users and the administrators.
Download the Microsoft Compute Cluster Pack software
development kit (SDK), if desired.
Dual Boot: Windows Compute Cluster Server 2003 and SUSELinux Enterprise Server 10
The information contained in this document represents the current view of Microsoft Corporation on the issues discussedas of the date of publication. Because Microsoft must respond to changing market conditions, it should not be interpretedto be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information presentedafter the date of publication.
This white paper is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, INTHISDOCUMENT.
Complying with all applicable copyright laws is the responsibility of the user. Without limiting the rights under copyright, nopart of this document may be reproduced, stored in, or introduced into a retrieval system, or transmitted in any form or byany means (electronic, mechanical, photocopying, recording, or otherwise), or for any purpose, without the express writtenpermission of Microsoft Corporation.
Microsoft may have patents, patent applications, trademarks, copyrights, or other intellectual property rights coveringbj t tt i thi d t E t l id d i itt li t f Mi ft th