Compute Cluster Administrator Operations
Applies To: Windows Compute Cluster Server 2003
Monitoring the cluster
You can view the health of your cluster at a glance by looking at the Start page of the Compute Cluster Administrator. This page displays node status such as number of nodes pending for approval or running jobs, processors in use, and useful information about the overall status of jobs submitted to the cluster.
In addition, the To Do List page displays notifications about the configuration of the cluster, such as how many RIS installation images are present and what networking interfaces are currently assigned.
Refer to the following section for a description of the cluster monitoring information available from the Compute Cluster Administrator:
Creating, adding, and removing nodes from a cluster
You can create a compute node and add it to a cluster using any of three methods:
Information about removing a node from a cluster can be found in the following section:
Node management includes operations on nodes such as running commands on a node or reimaging a node in case of failure. The following tasks describe node management operations:
The cluster administrator moves a new node through a distinct series of states. The administrator can only perform certain actions on a node if that node is in certain states. The following topic describes node states and administrator actions.
Monitoring cluster nodes
To monitor the state of the compute nodes in a cluster, see the following:
Managing RIS and RIS installation images
RIS installation images are used to install and configure compute nodes.
Managing users and administrators
Cluster users and cluster administrators are created and removed using the Manage Cluster User Wizard.
Managing job and job queue from Job Manager
The Compute Cluster Job Manager can be accessed directly from the Compute Cluster Administrator. To access the Job Manager, navigate to the Node Management page and click Launch Job Console. The Compute Cluster Job Manager will be displayed. For more information, see the following topics in the Compute Cluster Job Manager documentation.
View Jobs or Tasks
Submit a Job
Save a Job or a Task to a Template
Cancel a Job or Task
Requeue a Job or Task
For update purposes, Compute Cluster Server 2003 should be treated like any server network in an Active Directory domain. with two additional requirements:
Because Compute Cluster Pack is not part of the Windows operating system, all nodes must be joined to Microsoft Update in place of Windows Update in order to include Compute Cluster Pack updates. You can make this selection the first time you connect the nodes to Windows Update.
To avoid conflict with jobs, compute nodes should be idle and in the Paused state during update. The following steps are recommended.
Schedule updates for low usage times.
Notify users in advance of the shutdown.
Place all nodes to be updated in the Paused state.
After pausing the nodes, allow all running jobs to complete or cancel them for later requeue.
Perform the update.
Verify that update is complete and all necessary reboots have been performed.
Return the nodes to the Ready state.
As with any form of network-wide update, cluster update and the tasks surrounding it can be performed in a variety of ways, including stand alone scripts, Group Policy, and use of Windows Server Update. Other Microsoft network administrative and monitoring tools that can be used for updates are Microsoft Operations Manager 2005 and Microsoft Systems Management Server 2003.