Storagecentral 4.1 Cluster Support

WHITE PAPER

By David Shupp

March 2001


StorageCeNTral 4.1 Cluster Support

Contents

Introduction

Types of Clusters

Architecture

Installation

Conclusion

Resources

Introduction

StorageCeNTral 4.1 Advanced Server Edition supports clustered Windows NT/2000 Servers and meets IBM’s ClusterProven standard. Clustered servers provide high availability and load balancing by attaching two or more servers to a single shared disk array. This technical white paper discusses the architecture, installation, and operation of StorageCeNTral 4.1 Advanced Server Edition.

Types of Clusters

A Windows NT 4 clustered server can be installed with Windows NT 4 Enterprise Edition and Microsoft Cluster Server software. A Windows 2000 clustered server can be installed using Windows 2000 Advanced Server and Clustering Services. Two or more servers are attached to a single disk array using either SCSI or Fibre Channel. Other utilities which accomplish high availability through replication do not require StorageCeNTral 4.1 Advanced Server Edition. The Enterprise Edition may be used in these environments, and quotas can be replicated to the other replication server.

To support high availability, the shared cluster resources are online to one of the cluster servers at a time. This is called an Active/Passive cluster. In the event of a failure on the primary cluster node, the shared resources failover to the secondary node and users may continue to connect to the cluster without interruption.

To support load balancing and high availability, the shared cluster resources are online to both servers in the cluster at once. This is called an Active/Active cluster. This means each node in the cluster is the preferred owner of different resources in the shared array. Processing requests for these resources are shared by both servers, simultaneously. In the event of a failure of either cluster node, the shared resources on that node failover to the remaining cluster node and users continue to access these resources without interruption.

StorageCeNTral 4.1 Advanced Server Edition supports both Active/Passive and Active/Active types of clusters.

Architecture

StorageCeNTral 4.1 Advanced Server Edition consists of two components: QuotaAdvisor and DiskAdvisor. QuotaAdvisor is used to monitor and control storage. DiskAdvisor is used to audit storage using reports. Both components provide cluster support in several ways.

QuotaAdvisor Cluster Support

Quotas are stored on each partition attached to the server in a file called QAQuotasV4.dat. Threshold settings are also stored in two files called QAPolicyV4.dat and QAPolicyV4.idx. Quota actions are handled by the TruStor I/O filter driver called QAFilter. Quota management and threshold actions are handled by the QuotaAdvisor services: QAServer and QAPolicy. The driver and services are installed on each cluster server.

In the event of a failover, the cluster service on the secondary node detects the failure and issues a mount request for the shared cluster drives. QAFilter detects this mount request and begins managing quotas on that drive after it comes online. This is possible because the quota information is stored on the drive itself. The QAServer and QAPolicy services are restarted automatically and resume handling threshold actions on the remaining cluster node. No intervention by the Administrator is necessary.

Quota Templates are stored in the registry of each cluster server. If quotas are created automatically using “Learn Mode”, templates may be selected from the virtual server name of the “Cluster Group”. A StorageCeNTral cluster resource is created for each disk group. During a failover, this cluster resource copies quota templates from the primary cluster node to the remaining cluster node. This ensures that Learn Mode quotas continue to find the correct templates, regardless of which cluster node the resource is attached to.

Note: When creating and managing templates in a cluster, always use the templates on the node which contains the “Cluster Group” (quorum group). Do not create or modify templates from both node names, or you risk losing the templates on the receiving node if a failover occurs.


Example:

Cluster Groups-

Cluster Group

Disk Group 1

Disk Group 2

Quota templates should only be managed through the “Cluster Group”. This will prevent you from having two different versions of the same template on both nodes.

DiskAdvisor Cluster Support

DiskAdvisor reports can be scheduled to run on a regular basis. Reports are stored in the registry of each cluster server. Reports can be selected from either a management station or one of the virtual cluster server names. A StorageCeNTral cluster resource is created for each disk group. The purpose of this cluster resource is to copy DiskAdvisor reports from the primary cluster node to the secondary node during a failover. By doing this, you can be sure that reports that are scheduled on the virtual cluster server name continue to run, regardless of which node in the cluster owns the resource.

Installation

To install StorageCeNTral 4.1 Advanced Server Edition in a cluster, use the following steps:

1)  Ensure that your cluster meets the minimum requirements of StorageCeNTral 4.1 Advanced Server. (See the Release Notes on the install CD) Install the Exchange client software if you have chosen to use Exchange to send mail alerts.

2)  Prepare a service account for the StorageCeNTral services. The account should be a member of domain administrators for ease of setup. If a domain user account is used instead, you should either add this account to the local Administrators group on the server, or grant the account the “Increase Quotas” right.

3)  Obtain your mail configuration information. For SMTP, you require the SMTP server address (i.e. mail.yourdomain.com) and the sender address (). For Exchange, logon to the server once using the service account. The open Start>Settings>Control Panel>Mail. Click “Show Profiles” and create a profile name for the service account.

4)  Prepare to install on the first node by failing over all active resources to the other node. Once this node is “passive”, run StorageCeNTral 4.1 Advanced Server setup and restart the node.

5)  Once the first node has restarted, failover the second node to the first, making the second node the “passive” node. Run StorageCeNTral 4.1 Advanced Server setup on the second node and restart.

6)  Once the second node has restarted, failover the resources which are owned by this node (if needed).

Special Considerations

·  Use caution when failing over a cluster. Although a failover does not usually make the cluster unavailable, if the cluster service determines that a CHKDSK is needed, it may start the CHKDSK in the middle of failover and make the cluster unavailable until the CHKDSK completes. On a large array, a CHKDSK can take as long as an hour. For this reason, we recommend scheduling “down time” to install on a cluster just as you would for a stand-alone server.

·  When creating quotas or running reports on a clustered server, always select objects from the virtual server “Cluster Group”, not the name of the cluster node. This ensures that objects are found, regardless of which node in the cluster they are owned by. If you choose, you may want to add the “Cluster Group” virtual server names to your “Preferred Servers” list.

·  If you will be using Exchange to send mail, ensure that the Exchange client is installed on each node in the cluster, and the QAPolicy service account is logged in with an account that is associated with a mail profile on each node.

·  Use a management station to schedule reports. The schedule from one cluster node is not copied to the secondary node. This can be done by selecting the management station in the “Where To Execute” box on the Run tab.

Conclusion

StorageCeNTral 4.1 Advanced Server Edition provides support for clustered servers running Windows NT/2000. It meets the high availability demands of clustered servers by providing uninterrupted operation after cluster failover. It supports both Active/Passive and Active/Active cluster configurations and meets IBM’s ClusterProven standard.

Resources

http://support.wquinn.com

http://www.wquinn.com

http://www.pc.ibm.com/us/netfinity/clusterproven.html

http://www.microsoft.com/windows2000/library/howitworks/cluster/asoverview.asp

W. Quinn Associates, Inc.