Release Notes for Microsoft HPC Pack 2008 R2 Service Pack 1

Article
07/02/2015

Updated: February 2011

Applies To: Windows HPC Server 2008 R2

These release notes address late-breaking issues and information about Service Pack 1 (SP1) of Microsoft® HPC Pack 2008 R2. You can install HPC Pack 2008 R2 SP1 to upgrade an existing HPC cluster from Windows® HPC Server 2008 R2.

In these release notes:

Before you install HPC Pack 2008 R2 SP1
Compatibility of Windows HPC Server 2008 R2 and Windows HPC Server 2008 R2 SP1
Install Microsoft HPC Pack 2008 R2 SP1 on the head node
Install HPC Pack 2008 R2 SP1 on a high availability head node
Install HPC Pack 2008 R2 SP1 on compute nodes and broker nodes
Uninstall HPC Pack 2008 R2 with SP1
Restore the RTM version of HPC Pack 2008 R2 on the head node
Reinstall HPC Pack 2008 R2 with SP1 on the head node computer
Known issues

Before you install HPC Pack 2008 R2 SP1

Be aware of the following items and recommendations before you install HPC Pack 2008 R2 SP1 on the head node:

HPC Pack 2008 R2 SP1 is available only from the Microsoft Download Center (https://www.microsoft.com/downloads/en/details.aspx?FamilyID=aa86c849-15c0-4f97-a7b4-a23a2ba37890\&displaylang=en). The service pack is not made available automatically through Microsoft Update or through Windows Server Update Services.
When you install the service pack, new indexes and new parameters for some procedures are added to the HPC databases. To prevent potential data loss, create backups of the following databases before installing the service pack:
- Cluster management database
- Job scheduling database
- Reporting database
- Diagnostics database
You can use a backup method of your choice to back up the HPC databases. For more information and an overview of the backup and restore process, see Back Up and Restore in Windows HPC Server.
If the installation fails, you must completely uninstall HPC Pack 2008 R2, restore the HPC Pack 2008 R2 databases from a backup, and then follow the additional steps described in Restore the RTM version of HPC Pack 2008 R2 on the head node, later in this topic.
When you install the service pack, several settings related to HPC services are reset to their default values, including the following:
- Firewall rules
- Event log settings for all HPC services
- Service properties such as dependencies and startup options
- If the head node is configured for high availability in a failover cluster, the related resources that are configured in the failover cluster
Other details can be found in the following log file after you install the service pack: %windir%\temp\HPCSetupLogs\hpcpatch-DateTime.txt
Close all open windows for applications related to HPC Pack 2008 R2 before you apply the service pack. If you have windows open, such as HPC Cluster Manager, you may be prompted to reboot your computer to complete the installation.
Do not apply the service pack during critical times or while a long running job is still running. When you upgrade either a head node or a compute node in a cluster, you may be prompted to reboot the computer to complete the installation.
When you apply the service pack to your head node, the files that are used to deploy a compute node are also updated. If you install a new compute node or redeploy an existing compute node, the service pack is automatically applied to that compute node.

^ Top of page

Compatibility of Windows HPC Server 2008 R2 and Windows HPC Server 2008 R2 SP1

The following list summarizes the compatibility of Windows HPC Server 2008 R2 and Windows HPC Server 2008 R2 SP1 features:

HPC Cluster Manager in Windows HPC Server 2008 R2 SP1 can be used to manage a Windows HPC Server 2008 R2 head node. However, functionality such as new diagnostics tests that have been added to HPC Cluster Manager in Windows HPC Server 2008 R2 SP1 cannot be used in a Windows HPC Server 2008 R2 cluster. Doing so usually results in an error message. For information about features added in Windows HPC Server 2008 R2 SP1, see What's New in Windows HPC Server 2008 R2 Service Pack 1.
HPC Cluster Manager in Windows HPC Server 2008 R2 cannot be used to manage a Windows HPC Server 2008 R2 SP1 head node.
The job scheduler in Windows HPC Server 2008 R2 SP1 can be used to manage jobs in a Windows HPC Server 2008 R2 cluster.
The job scheduler in Windows HPC Server 2008 R2 can be used to manage jobs in a Windows HPC Server 2008 R2 SP1 cluster. However, the jobs cannot use job scheduling features that are new in Windows HPC Server 2008 R2 SP1.
In a cluster where the head node is upgraded to HPC Pack 2008 R2 SP1, compute nodes and Windows Communication Foundation (WCF) broker nodes should also be upgraded to HPC Pack 2008 R2 SP1.
HPC Pack 2008 R2 workstation nodes do not have to be upgraded to HPC Pack 2008 R2 SP1. However, user activity detection settings for workstation nodes that can be configured in Windows HPC Server 2008 R2 SP1 are ignored without warning on HPC Pack 2008 R2 workstations.

Note

A head node where HPC Pack 2008 R2 is installed cannot connect to workstation nodes where HPC Pack 2008 R2 SP1 is installed.

Note
A head node where HPC Pack 2008 R2 is installed cannot connect to workstation nodes where HPC Pack 2008 R2 SP1 is installed.

^ Top of page

Install Microsoft HPC Pack 2008 R2 SP1 on the head node

To install Microsoft HPC Pack 2008 R2 SP1 on the head node computer

Download the installation program for the 64-bit version of HPC Pack 2008 R2 SP1 (HPC2008R2_SP1_x64.exe) from the Microsoft Download Center (https://www.microsoft.com/downloads/en/details.aspx?FamilyID=aa86c849-15c0-4f97-a7b4-a23a2ba37890\&displaylang=en), and then save the service pack to installation media or to a network location.
Run HPC2008R2_SP1_x64.exe from the location where you saved the service pack.
Read the informational message that appears. If you are ready to apply the service pack, click OK.
Continue to follow the steps in the installation wizard.

Note
After you install the service pack, the computer restarts. You can confirm that HPC Pack 2008 R2 SP1 is installed on the head node. To view the version number in HPC Cluster Manager, on the Help menu, click About. If HPC Pack 2008 R2 SP1 is installed, the server version number and the client version number shown are similar to 3.1.xxxx.x.

^ Top of page

Install HPC Pack 2008 R2 SP1 on a high availability head node

If you have set up a head node for high availability in a failover cluster, use the following procedure to apply the service pack.

To install HPC Pack 2008 R2 SP1 on a high availability head node

Take the following high availability HPC services offline using Failover Cluster Manager: hpcscheduler, hpcsdm, hpcdiagnostics, and hpcsession.
Install the service pack on the active head node by running HPC2008R2_SP1_x64.exe.

After you install the service pack on the active head node, in most cases, the active head node restarts and fails over to the second head node.

Note

Because the second head node is not upgraded, Failover Cluster Manager might report a failed status for the resource group and the HPC services.
To install the service pack on the second head node, do the following:
1. Take the following high availability HPC services offline using Failover Cluster Manager: hpcscheduler, hpcsdm, hpcdiagnostics, and hpcsession.
2. Verify that the second head node is the active head node. If it is not, make the second head node the active head node using Failover Cluster Manager.
3. Install the service pack on the second head node.
  
  After you install the service pack on the second head node, the high availability HPC services are brought online automatically.

Note
Because the second head node is not upgraded, Failover Cluster Manager might report a failed status for the resource group and the HPC services.

Important
During the installation of the service pack on each head node configured for high availability, leave the SQL Server resources online.

^ Top of page

Install HPC Pack 2008 R2 SP1 on compute nodes and broker nodes

Note
When you apply the service pack to your head node, the files that are used to deploy a compute node or broker node from bare metal are also updated. After the head node is upgraded to SP1, if you install a new node or redeploy an existing node, HPC Pack 2008 R2 with SP1 is automatically installed on that node.

To install HPC Pack 2008 R2 SP1 on existing compute nodes and broker nodes, you can do one of the following:

Redeploy the nodes by using a method such as assigning a node template that reimages the node and installs HPC Pack 2008 R2 with SP1 using the installation files from the head node.
Run the HPC Pack 2008 R2 SP1 installation program (HPC2008R2_SP1_x64.exe) on individual nodes. For example you can copy HPC2008R2_SP1_x64.exe to a shared folder such as \\headnodename\SP1 and then access the existing compute nodes and broker nodes by Remote Desktop Connection to install the service pack from the shared folder.

Important
If you have WCF broker nodes that are configured for high availability in a failover cluster, you should install HPC Pack 2008 R2 SP1 on the high availability broker nodes as follows: Install HPC Pack 2008 R2 SP1 on the passive head node Fail over the active broker node to the passive broker node Install HPC Pack 2008 R2 SP1 on the passive head node (which has not yet been upgraded)

Note
For more information about how to perform Mayntenance on WCF broker nodes that are configured in a failover cluster, see Performing Mayntenance on WCF Broker Nodes in a Failover Cluster with Windows HPC Server 2008 R2.

Important
You cannot install HPC Pack 2008 R2 SP1 on compute nodes or on broker nodes by using either of the following methods: Adding the Apply Updates task to the Mayntenance phase of a node template Running a `clusrun` command to apply the service pack across a set of nodes in the cluster

^ Top of page

Uninstall HPC Pack 2008 R2 with SP1

Important
The installer for HPC Pack 2008 R2 SP1 does not support uninstallation back to HPC Pack 2008 R2. After you apply SP1, if you want to downgrade to the RTM version of HPC Pack 2008 R2, you must completely uninstall the HPC Pack 2008 R2 features from the head node computer and the other computers in your cluster. If you want, you can reinstall the RTM version of HPC Pack 2008 R2 and restore the data in the HPC databases.

The installer for HPC Pack 2008 R2 SP1 does not support uninstallation back to HPC Pack 2008 R2. After you apply SP1, if you want to downgrade to the RTM version of HPC Pack 2008 R2, you must completely uninstall the HPC Pack 2008 R2 features from the head node computer and the other computers in your cluster. If you want, you can reinstall the RTM version of HPC Pack 2008 R2 and restore the data in the HPC databases.

If you need to uninstall HPC Pack 2008 R2, uninstall the different features in the following order:

Microsoft HPC Pack 2008 R2 Services for Excel 2010
Microsoft HPC Pack 2008 R2 Server Components
Microsoft HPC Pack 2008 R2 Client Components
Microsoft HPC Pack 2008 R2 MS-MPI Redistributable Pack

Important
Not all features are installed on all computers. For example, Microsoft HPC Pack 2008 R2 Services for Excel 2010 is only installed on the head node when the HPC Pack 2008 R2 Enterprise and HPC Pack 2008 R2 for Workstation edition is installed.

When HPC Pack 2008 R2 is installed on the head node, other programs are installed with it. After uninstalling HPC Pack 2008 R2, you can remove the following programs if they will no longer be used:

Microsoft Report Viewer Redistributable 2008
Microsoft SQL Server 2008 (64-bit)

Note

This program also includes: Microsoft SQL Server 2008 Browser, Microsoft SQL Server 2008 Setup Support Files, and Microsoft SQL Server VSS Writer.
Microsoft SQL Server 2008 Native Client

Note
This program also includes: Microsoft SQL Server 2008 Browser, Microsoft SQL Server 2008 Setup Support Files, and Microsoft SQL Server VSS Writer.

Additionally, the following server roles and features might have been added when HPC Pack 2008 R2 was installed, and can be removed if they will no longer be used:

Dynamic Host Configuration Protocol (DHCP) Server server role
File Services server role
Network Policy and Access Services server role
Windows Deployment Services server role
Microsoft .NET Framework feature
Message Queuing feature

^ Top of page

Restore the RTM version of HPC Pack 2008 R2 on the head node

To restore the RTM version of HPC Pack 2008 R2 on the head node, follow these steps:

Uninstall all HPC Pack 2008 R2 components on the head node. For more information, see Uninstall HPC Pack 2008 R2 with SP1 , earlier in this topic.
Install the RTM version of HPC Pack 2008 R2 on the head node computer. Run setup.exe from the HPC Pack 2008 R2 installation media or from a network location.
If you are restoring head nodes that are configured for high availability in a failover cluster, do the following:
- Manually stop and disable only the following two services on each head node that are not managed by the failover cluster: hpcmanagement and hpcreporting. At an elevated command prompt window, type:
```
sc config hpcmanagement start= disabled
sc config hpcreporting start= disabled
net stop hpcmanagement
net stop hpcreporting
```
- Take the following high availability HPC services offline using Failover Cluster Manager: hpcscheduler, hpcsdm, hpcdiagnostics.

If you are restoring a head node that is not configured for high availability in a failover cluster, stop and disable all the HPC services. At an elevated command prompt window, type:

sc config hpcscheduler start= disabled
sc config hpcmanagement start= disabled
sc config hpcreporting start= disabled
sc config hpcsdm start= disabled
sc config hpcdiagnostics start= disabled
net stop hpcscheduler
net stop hpcmanagement
net stop hpcreporting
net stop hpcsdm
net stop hpcdiagnostics

Restore the HPC databases that you backed up before you applied the service pack. The restore method that you use will be determined by the database backup method you used.
Set the following registry key to indicate to the HPC Job Scheduler that it should enter restore mode when it restarts. At an elevated command prompt, type:
```
reg add HKLM\Software\Microsoft\HPC /v Restore /t REG_DWORD /d 1 /f
```
Important

Incorrectly editing the registry may severely damage your system. Before making changes to the registry, you should back up any valued data on the computer.
If you are restoring head nodes that are configured for high availability in a failover cluster, do the following:
- Bring the following high availability HPC services online using Failover Cluster Manager or a similar tool: hpcscheduler, hpcsdm, and hpcdiagnostics.
- Enable and start only the following two services on each head node that are not managed by the failover cluster: hpcmanagement and hpcreporting. At an elevated command prompt window, type:
```
sc config hpcmanagement start= auto
sc config hpcreporting start= auto
net start hpcmanagement
net start hpcreporting
```

Important
Incorrectly editing the registry may severely damage your system. Before making changes to the registry, you should back up any valued data on the computer.

If you are restoring a head node that is not configured for high availability in a failover cluster, enable and start all HPC services. At an elevated command prompt, type:

sc config hpcscheduler start= auto
sc config hpcmanagement start= auto
sc config hpcreporting start= auto
sc config hpcsdm start= auto
sc config hpcdiagnostics start= auto
net start hpcscheduler
net start hpcmanagement
net start hpcreporting
net start hpcsdm
net start hpcdiagnostics

Restart the HPC Management Service on all nodes other than the head node. You can use one of a variety of standard methods in the cluster to stop and restart the HPC Management Service on each node. For example, at an elevated command prompt on each node, type:
```
net stop HPCManagement
net start HPCManagement
```

Clean up any session data left in the broker nodes.

Log on to each broker node, and then on a HPC PowerShell session with elevated user rights, type:

System.Reflection.Assembly]::LoadWithPartialName("System.Messaging")
[System.Messaging.MessageQueue]::GetPrivateQueuesByMachine("localhost") | ? {"$($_.FormatName)" -like "*hpc*re*"} | % {[System.Messaging.MessageQueue]::Delete($_.Path)}

Note
If you have configured a broker node for high availability in a failover cluster, substitute the virtual broker node name for `“localhost”` in the PowerShell script. The process of deleting the session data can take a long time.

Additional considerations

You can use the Services Microsoft Management Console (MMC) snap-in to start or stop HPC services instead of running the net command.
After you have completed the restore, close and restart HPC Cluster Manager and HPC Job Manager.

^ Top of page

Reinstall HPC Pack 2008 R2 with SP1 on the head node computer

After you apply the service pack, if you need to on the head node and reapply SP1, while preserving the latest data in the HPC databases, follow the steps in Reinstalling HPC Pack 2008 R2 with Service Pack 1.

^ Top of page

Known issues

Installing HPC Pack 2008 R2 on the head node fails if the Windows Firewall service is stopped

If Windows Firewall service is stopped or turned off on the head node computer, installation of HPC Pack 2008 R2 does not complete successfully because the firewall rules required by Windows HPC Server 2008 R2 are not configured.

If this problem occurs during installation, entries such as the following may appear in the hpcMsi-DateTime.txt log file (located in the %Temp%\HPCSetupLogs folder):

CAQuietExec:  Firewall rule 'HPC Host (TCP-In)' addition failed: 0x800706d9
CAQuietExec:  Warning: failed to add rule HPC Host (TCP-In), continue
CAQuietExec:  Firewall rule 'HPC Host for controller (TCP-In)' addition failed: 0x800706d9
CAQuietExec:  Warning: failed to add rule HPC Host for controller (TCP-In), continue
CAQuietExec:  Firewall rule 'HPC Host x32 (TCP-In)' addition failed: 0x800706d9

Workaround

Ensure that the Windows Firewall service is started on the head node computer, and then try to install HPC Pack 2008 R2 again.

You can start the Windows Firewall service by using the Services MMC snap-in, or by running the following command in an elevated Command Prompt window:

net start MpsSvc

To verify the configuration of Windows Firewall on the head node computer, use the Windows Firewall with Advanced Security MMC snap-in.

Installation of SP1 may fail on a head node configured for high availability

On a Windows HPC Server 2008 R2 head node configured for high availability in a failover cluster, installation of HPC Pack 2008 R2 SP1 can fail if the head node restarts and fails over during the installation. This prevents the installation of the service pack from completing properly.

Workaround

To avoid this problem, it is recommended that, before you install the service pack on a head node configured for high availability, you manually take the high availability head node services offline by using Failover Cluster Manager. For more information about running the head node in a failover cluster, see Configuring Windows HPC Server 2008 R2 for High Availability of the Head Node (https://go.microsoft.com/fwlink/?LinkId=194691).

Windows PE failures may occur during bare metal deployment after SP1 is installed

After HPC Pack 2008 R2 SP1 is installed, if you attempt to deploy a node from bare metal, the deployment may time out in the Windows Preinstallation Environment (Windows PE).

Cause

In certain Windows HPC Server 2008 R2 SP1 clusters, the Windows PE execution client (ExecutionClient.exe) that was installed in Windows HPC Server 2008 R2 fails, causing bare metal deployment to time out.

Workaround

Run the UpdateHpcWinPE.cmd script on the head node to upgrade the Windows PE image (boot.wim. so that ExecutionClient.exe is updated for Windows HPC Server 2008 R2 SP1.

Important
To upgrade `boot.wim` on a head node configured for high availability, you must run the `UpdateHpcWinPE.cmd`script on each head node in the failover cluster,

To run the UpdateHpcWinPE.cmd script

Open an elevated Command Prompt window on the head node. Click Start, point to All Programs, click Accessories, right-click Command Prompt, and then click Run as administrator. If the User Account Control dialog box appears, confirm that the action it displays is what you want, and then click Continue.
Type the following command, and then press Enter:
```
UpdateHpcWinPE.cmd
```
The script performs the update of boot.wim.

You may experience errors or delays starting Windows Azure worker nodes, or worker nodes that are started may go into an error state

When starting Windows Azure worker nodes, the node health of some worker nodes may reMayn Transitional for an extended time period. Nodes that fail to start in Windows Azure after 1 hour will be unusable for the duration of the deployment (until the nodes are stopped and restarted).

Also, occasionally Windows Azure nodes that are successfully started may become unreachable or enter an error state. This may be due to Windows Azure service healing. When Windows Azure detects problematic nodes, service healing occurs automatically, moving role instances to new nodes. When this occurs, you might see your role instances restart automatically and go into an error state.

To determine if service healing has occurred

Sign in to the Windows Azure Portal (https://go.microsoft.com/fwlink/?LinkId=205529).
In the Manage Project view, click Download Support Information.
Review the information available, and then determine if a service healing event occurred during the time that you observed the worker nodes going into an error state.

Workaround

If you experience an error with worker nodes that were successfully started, you may need to take the nodes offline for the duration of the deployment.

Reporting of the availability of Windows Azure worker nodes differs from the information in Windows Azure Portal

Data about the availability of Windows Azure worker nodes that is stored in the HPCReporting database and that is reported in standard reports (or in custom reporting when data extensibility is enabled) does not exactly match the availability data that is shown in the Windows Azure Portal for the subscription. The availability data for the worker nodes that is stored in the HPCReporting database only approximates the actual availability of the worker nodes. For detailed availability and billing information about the worker nodes, refer to the data in the Windows Azure Portal for the subscription.

Cannot export information about Windows Azure worker nodes to a node XML file

In HPC Cluster Manager, in Node Management, if you select one or more Windows Azure worker nodes and then apply the Export Node XML File action, information about the Windows Azure worker nodes is not included in the exported XML file. Information about nodes other than Windows Azure worker nodes is exported properly. No warning message appears in HPC Cluster Manager when information about the Windows Azure worker nodes is not exported.

Windows Azure worker nodes added to a cluster using a cable modem may be unable to run jobs

If you add a large number of Windows Azure worker nodes to a Windows HPC Server 2008 R2 SP1 cluster that connects to the Internet using a cable modem, the nodes may deploy correctly but might be unable to run jobs. When attempting to run jobs on the worker nodes, you may see an error message similar to the following: The scheduler has experienced communication problem with the target node, and failed after 3 times.

Workaround

To temporarily resolve the problem so that the worker nodes can run jobs, try to reset the cable modem or the network adapter on the head node.

SOA sessions running on Windows Azure worker nodes may fail when the scheduler is configured in Balanced mode

Service-oriented architecture (SOA) sessions that are running on Windows Azure worker nodes are more likely to reach the message resend limits than sessions that are running on on-premises nodes, especially if the HPC Job Scheduler Service is running in Balanced mode.

Workaround

If you see messages fail on your Windows Azure worker nodes, try increasing the message resend limit (the default value is 3). This attribute can be set in the load balancing settings in the service configuration file (the <loadBalancing> element is in the <Microsoft.Hpc.Broker> section). For more information, see the broker settings section in SOA Service Configuration Files in Windows HPC Server 2008 R2.

HpcFile command does not support a trailing backslash in the destfile parameter when the value is surrounded by quotes

If you run the HpcFile command with destfile specified as anything other than the final parameter, and the value of the destfile parameter contains a trailing backslash (“\”) character and the value is surrounded by quotes, the command will fail with a parsing error.

Workaround

You can do one of the following:

Run the HpcFile command using a value of the destfile parameter that does not contain a trailing backslash character or does not require quotes
Run the HpcFile command specifying the destfile parameter in the final position on the command string

For more information about HpcFile command syntax, see HpcFile.

HpcFile command may fail to upload files to a virtual high availability head node name

If you run the HpcFile command to upload (write) files to a head node configured for high availability in a failover cluster, and you try to upload files to the virtual head node name, the command can fail if the head node fails over while the file copying is in progress. Files that are uploaded may appear to be “lost” to the client that is writing the files when the node fails over.

Workaround

To avoid this problem, configure a shared folder on a shared disk or file server in the failover cluster and write the files there instead of to the virtual cluster head node name.

For more information about HpcFile command syntax, see HpcFile.

Units are incorrect on the Disk Throughput chart

The units displayed on the Disk Throughput monitoring chart in HPC Cluster Manager are incorrect. The units shown on the chart are Kbytes/sec. The units should be Bytes/sec.

Workaround

You can correct the units displayed on the Disk Throughput monitoring chart by following this procedure.

To correct the units displayed on the Disk Throughput chart

Start HPC PowerShell with administrative privileges. On the head node, click Start, point to All Programs, click Microsoft HPC Pack 2008 R2, right-click HPC PowerShell, and then click Run as administrator.

Run the following PowerShell cmdlet:

Export-hpcmetric –path:c:\tests\metric.xml

Open the file metric.xml in a text editor such as Notepad.
In metric.xml, locate and change “Kbytes” to “Bytes”, and then save the file.

Run the following PowerShell cmdlet:

Import-hpcmetric –path:c:\tests\metric.xml

Installing .NET Framework 4.0 on a broker node will stop the HPC Broker Service

After .NET Framework 4.0 is installed on a broker node, the node restarts automatically but the HPC Broker Service on the node is not restarted. You may see an error message similar to: Broker Service is Unreachable. The node does not perform properly as a broker node.

Workaround

To work around this problem, after installing .NET Framework 4.0, manually start the HPC Broker Service on the broker node. Alternatively, you can first disable the broker node role on the node (changing the functionality of the node to a compute node), and then re-enable the broker node role.

To disable and enable the broker role node

In HPC Cluster Manager, in Node Management, take the broker node offline.
Right-click the node and then click Change Role.
In the Change Role dialog box, unselect (remove the check from) the WCF Broker Node checkbox, make sure that that the Compute Node checkbox is selected, and then click OK.
Right-click the node and then click Change Role.
In the Change Role dialog box, select the WCF Broker Node checkbox, select the Compute Node (if desired), and then click OK.

GPGPU jobs may not run on workstation nodes that have user activity detection settings configured in the availability policy

General purpose computations on graphics processor units (GPGPU) programs may not run on workstation nodes for which user activity detection settings are configured in the workstation node availability policy. When attempting to run GPGPU jobs, you may see the jobs starting and then immediately getting canceled. When the job is canceled, the workstation node may not run further jobs depending on the user activity detection settings.

Cause

This problem can occur if the availability policy configured in the workstation node template includes a CPU usage percentage below which the node can run jobs. When the workstation prepares to run a GPGPU job a high CPU load is placed on the workstation that exceeds the CPU usage percentage configured in the availability policy.

Note
This problem can occur only if the job is set with job environment variable HPC_CREATECONSOLE=True. This problem does not occur if the job is set with the job environment variable HPC_ATTACHTOSESSION or HPC_ATTACHTOCONSOLE.

Workaround

To run GPGPU jobs on workstation nodes, do not enable user activity detection settings in the availability policy of the workstation node template.

NullReferenceException when taking broker node offline during creation of a SOA session

If a broker node is taken offline during the creation of a service-oriented architecture (SOA) session, a NullReferenceException may occur. The SOA session is not created properly.

Workaround

To work around this problem, the SOA client can retry creating the SOA session.

BrokerResponse.GetUserData may throw NullReferenceException

A NullReferenceException might be thrown when calling BrokerResponse.GetUserData<T> if SetResponseHandler is used.

Workaround

To work around this problem, call BrokerResponse.GetUserData after calling BrokerResponse.Result.

Share via

Release Notes for Microsoft HPC Pack 2008 R2 Service Pack 1

Before you install HPC Pack 2008 R2 SP1

Compatibility of Windows HPC Server 2008 R2 and Windows HPC Server 2008 R2 SP1

Install Microsoft HPC Pack 2008 R2 SP1 on the head node

To install Microsoft HPC Pack 2008 R2 SP1 on the head node computer

Install HPC Pack 2008 R2 SP1 on a high availability head node

To install HPC Pack 2008 R2 SP1 on a high availability head node

Install HPC Pack 2008 R2 SP1 on compute nodes and broker nodes

Uninstall HPC Pack 2008 R2 with SP1

Restore the RTM version of HPC Pack 2008 R2 on the head node

Reinstall HPC Pack 2008 R2 with SP1 on the head node computer

Known issues

Installing HPC Pack 2008 R2 on the head node fails if the Windows Firewall service is stopped

Workaround

Installation of SP1 may fail on a head node configured for high availability

Workaround

Windows PE failures may occur during bare metal deployment after SP1 is installed

Cause

Workaround

To run the UpdateHpcWinPE.cmd script

You may experience errors or delays starting Windows Azure worker nodes, or worker nodes that are started may go into an error state

To determine if service healing has occurred

Workaround

Reporting of the availability of Windows Azure worker nodes differs from the information in Windows Azure Portal

Cannot export information about Windows Azure worker nodes to a node XML file

Windows Azure worker nodes added to a cluster using a cable modem may be unable to run jobs

Workaround

SOA sessions running on Windows Azure worker nodes may fail when the scheduler is configured in Balanced mode

Workaround

HpcFile command does not support a trailing backslash in the destfile parameter when the value is surrounded by quotes

Workaround

HpcFile command may fail to upload files to a virtual high availability head node name

Workaround

Units are incorrect on the Disk Throughput chart

Workaround

To correct the units displayed on the Disk Throughput chart

Installing .NET Framework 4.0 on a broker node will stop the HPC Broker Service

Workaround

To disable and enable the broker role node

GPGPU jobs may not run on workstation nodes that have user activity detection settings configured in the availability policy

Cause

Workaround

NullReferenceException when taking broker node offline during creation of a SOA session

Workaround

BrokerResponse.GetUserData may throw NullReferenceException

Workaround

Additional resources