Adding Unmanaged Servers to a Windows HPC Cluster
Updated: January 13, 2014
Applies To: Microsoft HPC Pack 2012, Microsoft HPC Pack 2012 R2
Using a Windows HPC cluster that is running at least HPC Pack 2008 R2 with Service Pack 3 (SP3), you can harvest extra processing cycles from other servers on your network that are running the Windows Server operating system. The other servers can be used opportunistically to run cluster jobs during periods of low demand for their typical network workloads, such as running a file server or print server. In addition, the servers offer the cluster administrator greater flexibility in managing cluster users, because the user accounts that can access the servers are managed separately from the HPC cluster users and HPC cluster administrators that are configured to access the other cluster nodes. For example, unlike the local Administrators group on a head node or a dedicated compute node in the cluster, the local Administrators group on a network server is not automatically populated with the same user accounts that are added as HPC cluster administrators. In the context of the HPC cluster, these servers are considered “unmanaged”.
This topic summarizes the requirements and steps for adding unmanaged servers for cycle harvesting to your Windows HPC cluster. The process for adding unmanaged servers in the cluster is generally similar to the steps for adding workstation nodes, with certain differences. Links are provided from this topic to the related steps in the Adding Workstation Nodes to a Windows HPC Cluster. Like workstation nodes, unmanaged servers can automatically be brought online (become available to run cluster jobs) according to a weekly availability policy—for example, every night on weekdays and all day on weekends—or manually.
In this topic:
The head node computer (or computers) where HPC Pack is installed must be fully configured (that is, all the steps required in the Deployment To-do List have been completed). The head node must be running at least HPC Pack 2008 R2 with SP3.
The unmanaged servers must be running a version of the Windows Server operating system that is supported by your version of HPC Pack, as listed in the following table. It is recommended that you apply the latest service pack of the Windows Server operating system.
Version of HPC Pack
Operating system requirement for unmanaged server nodes
HPC Pack 2012
HPC Pack 2008 R2 (with at least SP3)
Windows Server 2008 R2 Standard, Windows Server 2008 R2 Enterprise, Windows Server 2008 R2 Datacenter, or Windows Server 2008 R2 HPC Edition
You mush have installation media for HPC Pack 2008 R2 with SP3 or a later version of HPC Pack.
If you will be using HPC Pack 2008 R2 with SP3 or SP4, note the following:
Beginning with HPC Pack 2008 R2 with SP3, the edition previously known as HPC Pack 2008 R2 Enterprise and Workstation was renamed HPC Pack 2008 R2 Enterprise and HPC Pack 2008 R2 for Cycle Harvesting edition. Installation on an unmanaged server requires new HPC Pack 2008 R2 Enterprise and HPC Pack 2008 R2 for Cycle Harvesting media to be created by either downloading the new disc from your volume license website or by downloading the appropriate media integration package to upgrade the release-to-manufacturing (RTM) version of HPC Pack 2008 R2.
To enable HPC job submission to the unmanaged servers, you must ensure that a user account that is a member of the Users group on each server is added to the cluster with a role that has permissions to submit cluster jobs. You can do this by using HPC Cluster Manager or by using HPC PowerShell. For more information about the user accounts, see “Considerations for user accounts used on the workstation computers” in Requirements for Adding Workstation Nodes with Microsoft HPC Pack, and Security Considerations for Designating HPC Cluster Users and Administrators.
It is strongly recommended that you do not add accounts that have local administrative privileges on the servers. If you do this, the users who have local administrative privileges can affect the performance and the outcome of HPC jobs that are running on the servers, or the ability of the servers to carry out their default workloads. It might also be possible for an attacker who has local administrative privileges to create a service that can affect HPC services or jobs that are running on the server or on other nodes in the cluster.
Servers that you add to your HPC cluster for cycle harvesting are generally on the enterprise network. In most cases, you should configure a cluster network topology that includes the enterprise network: Topology 2, 4, or 5. For information about choosing a network topology to connect workstations or unmanaged servers to the cluster and the impact of the topology on cluster performance and functionality, see Appendix 1: Cluster Network Topologies for Workstation Nodes in the Adding Workstation Nodes Step-by-Step Guide.
To add unmanaged servers to your HPC cluster, you must first create a node template. To do this, follow the general steps in Step 1: Create a Workstation Node Template in the Adding Workstation Nodes Step-by-Step Guide. To add unmanaged servers, like workstation nodes, you use the workstation node and unmanaged server node template. The node template includes settings that you can configure to make the unmanaged server nodes available to run jobs manually, or automatically according to an availability policy that you configure. The availability policy that you can configure in the node template can include a weekly availability schedule and, optionally, settings to detect user activity on the computers.
Because unmanaged servers usually do not have a logged-on user, only the weekly schedule (if configured) applies to unmanaged servers. User activity detection settings generally are ignored, even if they are configured in the template.
You must install HPC Pack on each unmanaged server that you want to add to the cluster. In general you should install the same version of HPC Pack that is installed on the head node of the cluster. To do this, follow the general procedure (using the appropriate installation media) that is described in Step 2: Install HPC Pack on the Workstation Computers in the Adding Workstation Nodes Step-by-Step Guide. The method that is used to install HPC Pack depends on the management of the server infrastructure, as well as security policies and deployment practices in your organization.
If you choose to install HPC Pack manually, you can use the HPC Pack Installation Wizard or a command-line installation, including unattended installation. For a command-line installation, you can use the same command line and options that are used to install HPC Pack on a workstation computer.
To assign a node template to one or more unmanaged servers, follow the general steps in Step 3: Assign a Workstation Node Template in the Adding Workstation Nodes guide. Assigning a node makes the servers available to run jobs according to the manual or automatic availability policy that is configured in the node template. HPC Cluster Manager and the HPC PowerShell cmdlets can also be used to monitor the cluster operations on the servers and to perform cluster diagnostic tests.
By default, unmanaged servers that are added to the cluster appear in the UnmanagedServerNodes group in the Node Management pane of HPC Cluster Manager. You can add the servers to other node groups in the same way that you add other cluster nodes to node groups. For more information, see Grouping Nodes.
After the node template is assigned, the unmanaged servers can be brought online to run cluster jobs, and then taken offline to resume their normal workloads. Nodes that are configured in the node template to be brought online and offline manually will initially be offline. The workstation nodes that are configured to be brought online and offline according to a weekly availability policy will begin to follow that policy, and they will be brought online automatically during the scheduled intervals.
To bring online the workstation nodes that are configured to be brought online and offline manually, use the following procedure.
To bring unmanaged servers online manually to run jobs
In HPC Cluster Manager, in Node Management, in the Navigation Pane, click Nodes.
In the views pane, select one or more servers.
Right-click the selected servers, and then click Bring Online (or click Bring Online in the Actions pane).