共用方式為

Migrating a Failover Cluster Running Windows HPC Server 2008 to Windows HPC Server 2008 R2 Step-by-Step Guide

  • 發行項

Updated: August 2011

Applies To: Windows HPC Server 2008 R2

This guide describes the process of migrating a failover cluster that supports the head node of a high-performance computing (HPC) cluster. You start the migration with a failover cluster that contains two servers running Windows Server® 2008, preferably with the latest service packs and software updates, and Windows® HPC Server 2008 with Service Pack 2 (SP2).

Important
  • Before you can begin the migration that is described in this guide, both servers in the old failover cluster must be running Windows HPC Server 2008 with SP2. Windows HPC Server 2008 SP1 is a prerequisite for Windows HPC Server 2008 SP2. To check the version of Windows HPC Server 2008, in HPC Cluster Manager, on the Help menu, click About. The Server version for Windows HPC Server 2008 Service Pack 2 is 2.2.1841.0.

  • The procedures in this guide are intended only for migrating the old failover cluster to a new failover cluster that supports a Windows HPC Server 2008 R2 head node. They are not intended for migrating to a single Windows HPC Server 2008 R2 head node or for an in-place migration of the failover cluster on the existing computers.

Running a head node in a failover cluster can increase the availability of the head node. Because the failover cluster contains two servers that work together, if there is a failure of the server that is acting as the head node, the other server in the failover cluster automatically begins acting as the head node (in a process known as failover). Even if an outage occurs on a server, existing jobs can continue to run and new jobs can be submitted.

During the migration described in this guide, we recommend that you perform installations and migrate settings onto three or four new servers:

  • Server or servers to run SQL Server: This can be one server that is running Windows Server 2008 R2 with SQL Server 2008 SP1 (or later), or two servers in a failover cluster running the same software.

  • Servers to run HPC Server 2008 R2: Two new servers that run the Windows Server 2008 R2 operating system with Windows HPC Server 2008 R2. These servers are configured as a failover cluster that supports the head node services for the new HPC cluster.

The migration process described in this guide migrates most, but not all of the information on the old cluster:

  • Information that is migrated: HPC cluster configuration data and job history.

  • Information that is not migrated: Operation history, monitoring history (such as charts and performance counters), and results of diagnostic tests.

The following sections provide details about migrating a failover cluster that supports the head node of an HPC cluster:

Step 1: Set up and configure the new servers to which you will migrate

Step 2: Migrate resource settings from the failover cluster that supports the old head node

Step 3: Prepare to export configuration data and then run the export command

Step 4: Take the old head node offline and bring the migrated clustered instance online

Step 5: Install HPC Pack 2008 R2 on the new failover cluster and import configuration data

Step 6: Confirm that the migrated head node can fail over correctly

Step 1: Set up and configure the new servers to which you will migrate

The following procedure lists steps from a document that describes how to create a head node running Windows HPC Server 2008 R2 (and Windows Server 2008 R2) in a failover cluster.

To set up and configure the new servers to which you will migrate

  1. Perform the steps in the following topics:

    1. Install Windows Server 2008 R2 on Multiple Servers (https://go.microsoft.com/fwlink/?LinkId=201566).

    2. Set Up Shared Storage for Servers for the Head Node (https://go.microsoft.com/fwlink/?LinkId=201567).

    3. Set Up Failover Clustering and File Services for Servers that Will Run the Head Node (https://go.microsoft.com/fwlink/?LinkId=201568).

    4. Set Up Remote SQL Server 2008 SP1 for the HPC Cluster (https://go.microsoft.com/fwlink/?LinkId=201569).

      This step includes creating four databases. We suggest that you use the following standardized names to simplify the process later:

      • HPCManagement

      • HPCScheduler

      • HPCReporting

      • HPCDiagnostics

    Important
    Do not install HPC Pack 2008 R2 yet on the new servers. You must migrate configuration information to those servers before you install HPC Pack 2008 R2.

Step 2: Migrate resource settings from the failover cluster that supports the old head node

The following procedure describes how to migrate resource settings from the old failover cluster, that is, the failover cluster that ran the head node of the old HPC cluster. Those resources make up the “clustered instance” that runs the head node. (A clustered instance is the unit that fails over from one server to the other in a failover cluster). This procedure prepares you for later procedures in which you modify the migrated clustered instance and then install HPC Pack 2008 R2 on the new failover cluster.

Important
Before you begin migrating resource settings from the old failover cluster, make sure that you have installed Windows HPC Server 2008 SP2 on both servers in the old failover cluster. For information about determining which version of Windows HPC Server 2008 you are running, see the important note at the beginning of this guide.

To migrate resource settings from the failover cluster that supports the old head node

  1. On the new failover cluster, that is, the failover cluster that you created on servers running Windows Server 2008 R2, click Start, click Administrative Tools, and then click Failover Cluster Manager. If the User Account Control dialog box appears, confirm that the action it displays is what you want, and then click Yes.

  2. If the new failover cluster is not displayed, in the console tree, right-click Failover Cluster Manager, click Manage a Cluster, and then select or specify the cluster that you want.

  3. Under Configure, click Migrate services and applications.

  4. Read the first page of the Migrate a Cluster Wizard, and then click Next.

  5. Specify the name or IP address of the old failover cluster, or of a server in the old failover cluster, and then click Next.

    In the next page of the wizard, if you click View Report and read the report, you will see that a number of the resources in the old cluster are not eligible for migration. This is expected. You will complete the migration of the old clustered instance in later procedures in this guide. (The wizard also provides a report after it finishes.)

  6. Follow the instructions in the wizard to select the clustered instance of the head node in the old HPC cluster. (A clustered instance is also called a resource group.)

    Figure1   Select Resource Groups page in the Migrate a Cluster Wizard

  7. Follow the instructions in the wizard to specify a disk in new storage that will be used by the head node of the new HPC cluster.

    Note
    You cannot use the old storage for the new failover cluster.

    Figure 2  Select Storage page in the Migrate a Cluster Wizard

  8. Follow the instructions in the wizard to perform the migration. After the Summary page appears, click View Report to confirm that the clustered instance of the head node migrated successfully; however, do not bring it online yet.

Step 3: Prepare to export configuration data and then run the export command

The following procedure describes how to export configuration data that is used by the old HPC cluster so that it can later be imported for use by the new HPC cluster.

To prepare to export configuration data and then run the export command

  1. Collect the following information, which you will need when you run a command to export configuration data from the head node of the old HPC cluster to the head node of the new HPC cluster:

    • A disk drive, folder, or network path that is accessible from the old head node, from which you can run Setup for HPC Pack 2008 R2. Setup contains the export command that you will use to export configuration data.

    • The name of the old SQL Server failover cluster (the one used by the head node). Within Failover Cluster Management, this name is displayed as a name resource within the clustered instance of the old head node. The name identifies the SQL Server failover cluster on the network (it is not the name of a physical server). In a later step, you will specify this name followed by \computecluster to specify the database that is used by the old head node.

  2. In the old HPC cluster, allow existing jobs to finish, or cancel existing jobs, and then take all compute nodes offline.

  3. On a server in the old failover cluster, take note of which server is currently running the clustered instance of the head node. On that server, confirm the following items:

    • In Failover Cluster Management, confirm that the clustered instance of the head node is online (specifically, that the Microsoft HPC Scheduler Service and the Microsoft HPC SDM Store Service are online).

    • In Server Manager, expand Configuration, click Services, and confirm that the HPC Management Service is started.

    • Confirm that a doMayn controller is available.

    Note
    The export command will not run if the appropriate services are not running and a doMayn controller is not available.

  4. On the same server in the old failover cluster, create an empty folder.

    Note
    Create the folder on the server itself, not in the context of the failover cluster. You use this folder for a short time as a place to hold the data that will be exported from the old head node and imported into the new head node.

  5. On the same server in the old failover cluster, open a Command Prompt window with elevated administrative rights: click Start, right-click Command Prompt, and then click Run as administrator.

  6. Type the following command:

    <HPCSetupLocation>\Setup\export.cmd <ExportFolder> <OldSQLServerName>\computecluster

    where <HPCSetupLocation> is the DVD drive or network location of the setup media for HPC Pack 2008 R2, <ExportFolder> is the full path of the export folder that you created previously, and <OldSQLServerName>\computecluster specifies the database name of the SQL Server database that supports the old HPC cluster. (In this database name, <OldSQLServerName> is the name that identifies the SQL Server failover cluster on the network and computecluster is the instance name of the SQL Server database on the old head node.)

    This command makes a copy of a set of configuration information that you will later import to the new head node (in the new failover cluster).

    Note
    The export command might change the Windows PowerShell execution policy to RemoteSigned. As needed, adjust the execution policy after the export command completes. For information about Windows PowerShell execution policies, see Get-ExecutionPolicy (https://go.microsoft.com/fwlink/?LinkId=113315) and Set-ExecutionPolicy (https://go.microsoft.com/fwlink/?LinkId=113394).

  7. To confirm that the export command succeeded, view the contents of the folder that you created in step 4. After a successful export, this folder will contain a subfolder called Backup (with additional subfolders and files), and at least one file with a file name extension of .log.

Step 4: Take the old head node offline and bring the migrated clustered instance online

The following procedure describes actions that are necessary before you can install HPC Pack 2008 R2 in the new failover cluster. First you must take the head node for the old HPC cluster offline and perform several intermediate steps. Then you will bring the migrated clustered instance online. (The migrated cluster instance in the new failover cluster will eventually become the new head node.)

To take the old head node offline and bring the migrated clustered instance online

  1. In the old failover cluster, click Start, click Administrative Tools, and then click Failover Cluster Management. Falls das Dialogfeld Benutzerkontensteuerung angezeigt wird, bestätigen Sie, dass die angezeigte Aktion der gewünschten Aktion entspricht, und klicken Sie anschließend auf Weiter.

  2. If the cluster that you want to manage is not displayed, in the console tree, right-click Failover Cluster Management, click Manage a Cluster, and then select or specify the cluster you want.

  3. If the console tree is collapsed, expand the tree under the old failover cluster.

  4. Under Services and Applications, expand the console tree.

  5. Right-click the clustered instance of the head node, and then click Take this service or application offline. If prompted, confirm your action.

  6. On a server in the old failover cluster, click Start, click Administrative Tools, and then click Services. (Falls das Dialogfeld Benutzerkontensteuerung angezeigt wird, bestätigen Sie, dass die angezeigte Aktion der gewünschten Aktion entspricht, und klicken Sie anschließend auf Ja.)

  7. In the list of services, view each of the following services, and if the service is not Disabled, right-click it, click Properties, click the Stop button and then, in the list for Startup type, click Disabled. These are the services to disable:

    • DHCP Server

    • Windows Deployment Services Server

    • Routing and Remote Access

      Note
      After you disable Routing and Remote Access, any compute nodes that depend on this service for connectivity will no longer have access to resources (such as doMayn controllers) that are on your Enterprise network.

  8. Repeat steps 6-7 on the other server in the old failover cluster.

  9. In the new failover cluster, click Start, click Administrative Tools, and then click Failover Cluster Manager. Falls das Dialogfeld Benutzerkontensteuerung angezeigt wird, bestätigen Sie, dass die angezeigte Aktion der gewünschten Aktion entspricht, und klicken Sie anschließend auf Ja.

  10. In the new failover cluster, in the console tree, right-click Failover Cluster Manager, click Manage a Cluster, and then select or specify the cluster that you want to manage.

  11. Under Services and Applications, expand the console tree, and click the clustered instance that you migrated. Make sure that the list of resources in the clustered instance is displayed. The following screenshot shows a migrated clustered instance that includes a Fulltext resource (some migrated clustered instances will not include this resource).

    Figure 3  Example of a migrated clustered instance in Failover Cluster Manager

  12. In the center pane, locate the File Server resource and, if present, the SQL Server Fulltext resource. Right-click each of these resources (if present), and then click Delete. (These resources are no longer needed.)

    Important
    For this action, we recommend that you carefully right-click the resource or resources, rather than using the Action menu, to be sure that you delete the resource, and not the clustered instance.

  13. In the console tree, click the clustered instance that you migrated and view the Current Owner (in the center pane summary). Confirm that you are working in the server that currently owns the migrated clustered instance. If you are not, either go to the other server in the new failover cluster, or right-click the migrated clustered instance, click Move this service or application to another node, click the available choice of node, and when prompted, confirm your choice.

    Also, in the clustered instance, in the center pane under Disk Drives, expand the cluster disk and take note of the drive letter that it uses.

  14. With the clustered instance still selected (in the console tree), in the Actions pane, click Bring this service or application online.

  15. On the same server in the new failover cluster, open a Command Prompt window with elevated administrative rights: click Start, right-click Command Prompt, and then click Run as administrator.

  16. Type the following command:

    net share <sharename>=<drive>:\

    where sharename is the name of a share that does not already exist on the server, and drive is the drive letter of the cluster disk that is used by the migrated clustered instance.

    The new shared folder appears in Failover Cluster Manager, in the migrated clustered instance, under Shared Folders. In the following screenshot, the shared folder is called share1.

    Figure 4  Example of a migrated clustered instance with modifications

Step 5: Install HPC Pack 2008 R2 on the new failover cluster and import configuration data

The following procedure describes how to install HPC Pack 2008 R2 on the new failover cluster and then import the configuration data that you previously exported from the old head node.

To install HPC Pack 2008 R2 on the new failover cluster and import configuration data

  1. For the new, remote SQL Server instance (running SQL Server 2008 SP1 or later) that you created in Step 1: Set up and configure the new servers to which you will migrate, review the server name and instance name. You will need this information when you install HPC Pack 2008 R2.

  2. In the new failover cluster, if Failover Cluster Manager is not already started, start it. Under Services and Applications, click the migrated clustered instance to see which server currently owns it.

  3. On the server that currently owns the migrated clustered instance, install HPC Pack 2008 R2, as described in Install HPC Pack 2008 R2 on a Server that Will Run Head Node Services (https://go.microsoft.com/fwlink/?LinkId=201570).

    Important
    • The current version of HPC Pack 2008 R2 is HPC Pack 2008 R2 with Service Pack 2. It is recommended that you install HPC Pack 2008 R2 with Service Pack 2 using integrated (or slipstream) installation media.

    • Do not perform any actions in the Deployment To-do List. The process of importing data from the old cluster will perform some of these actions for you.

  4. In Failover Cluster Manager, confirm that the migrated clustered instance is still owned by the first server on which you installed HPC Pack 2008 R2. (As needed when performing the next few steps, confirm that the migrated clustered instance is still owned by the first server.) Then go to the other server in the new failover cluster and install HPC Pack 2008 R2, as described in Install and Configure HPC Pack 2008 R2 on the Other Server that Will Run Head Node Services (https://go.microsoft.com/fwlink/?LinkId=201572).

  5. Share the export folder that you created on a server on the old failover cluster, that is, share the folder that was specified in the export command by <ExportFolder>. When you share the folder, give the following accounts Contributor (read and write) permissions:

    • The account that you will use when you import the data that you exported earlier.

    • The SQL Server service account that you specified when you created the remote SQL Server instance for the new HPC cluster.

  6. To confirm that your permissions allow you to view the shared export folder from another server, go to a server in the new failover cluster, log on with the account that you will use when you import the data for the head node, open a Command Prompt window, and type:

    net view \\<ServerThatContainsExportFolder>

    where <ServerThatContainsExportFolder> is the name of the server that hosts the shared export folder. The display should show the name of the shared folder.

  7. In the remote SQL Server instance that you created for the new HPC cluster, close all data connections for the four databases. This includes closing SQL Server Management Studio if it is open and connected to the remote SQL Server instance.

  8. On the first server in the new failover cluster (which should still be the owner of the migrated clustered instance), open a Command Prompt window with elevated administrative rights: click Start, right-click Command Prompt, and then click Run as administrator.

  9. Type the following command:

    <HPCSetupLocation>\Setup\import.cmd \\<ServerThatContainsExportFolder>\<ExportFolder>

    where <HPCSetupLocation> is the DVD drive or network location of the setup media for HPC Pack 2008 R2 and \\<ServerThatContainsExportFolder>\<ExportFolder> is the network path to the shared export folder. (This is located on a server in the old failover cluster.)

  10. When you are prompted, press Y to confirm that the path to the export folder is accessible.

    Configuration data is imported to the new failover cluster.

    Important
    If a problem occurs during the import of the configuration data, determine the cause from the command output, or view the log files in the \\<ServerThatContainsExportFolder>\<ExportFolder>\Backup folder. After you resolve the problem, run import.cmd again. If the HPC services fail to start, you should restore the databases manually from the following location: \\<ServerThatContainsExportFolder>\<ExportFolder>\Backup\Database. Then, run import.cmd again to complete the import of the configuration data.

  11. On a server in the new failover cluster, open HPC Cluster Manager.

  12. In the To-do List, perform the first and second steps. (The other steps were already performed when you ran the import command.)

    For more information about these steps, see Configure the Head Node on the First Server (https://go.microsoft.com/fwlink/?LinkId=201571).

  13. If the old HPC cluster used a custom power control script, as described in Appendix 5: Scriptable Power Control Tools (https://go.microsoft.com/fwlink/?LinkId=202856), in the new failover cluster, copy the script from one server to the other server as follows:

    1. On the first server on which you installed HPC Pack 2008 R2 (the one on which you ran the import command), navigate to the Bin folder of the installation path for HPC Pack 2008 R2. For example, with the default path, the Bin folder is here:

      C:\Program Files\Microsoft HPC Pack\Bin\CcpPower.cmd

    2. Copy CcpPower.cmd to the same location (the Bin folder of the installation path for HPC Pack 2008 R2) on the other server in the new failover cluster.

  14. Stop sharing the folders that you shared earlier, as follows:

    • On the server with the shared export folder, stop sharing the export folder.

    • On a server in the new failover cluster, in Failover Cluster Manager, in the migrated clustered instance under Shared Folders, right-click the shared folder that you created with the net share command, and then click Stop Sharing.

    Note
    This step is a best practice to prevent unintended access to these folders.

  15. As appropriate in the new HPC cluster, image compute nodes, Windows Communication Foundation (WCF) broker nodes, or both by using the templates that were imported with the rest of the configuration data.

Step 6: Confirm that the migrated head node can fail over correctly

To confirm that the migrated head node can fail over correctly, use the procedure in Validate Installation of Head Node Services in the Context of the Failover Cluster (https://go.microsoft.com/fwlink/?LinkId=201573).