Article ID: 171277 - View products that this article applies to.
This article was previously published under Q171277
If a failover condition occurs within Microsoft Cluster Server (MSCS), resources within the failing group will transition to another node within the cluster. The time required to bring resources online may vary based on resource types and other factors.
Microsoft Cluster Server allows for resources to be members of a group of like resources. Resources may run on only one node at a time. Groups are the basic failover units within a cluster and may be owned by only one system at a time. If a resource failure occurs within a group, the resource manager may try to bring the resource online or failover the entire group to another node in the cluster. During failover, clients using cluster resources may experience little or no change in activity, or else may need to reconnect. Client impact depends greatly on the type of resources they use and other resource factors.
The time required to bring a resource online may depend on the following factors:
Resources may be configured to stay offline after a specified number of failures. They may also be restricted from running on specific nodes within the cluster. Use Cluster Administrator to modify these properties.
Groups may have a preferred owner and may be configured to failback automatically if the preferred owner node is available. Groups may be configured not to failback to another node or to failback only during a specific time window.
Resources may also be dependent on other resources within their group. As a result, the dependent resources will not come online until the resources they depend upon are online.
Additionally, if a node fails, surviving cluster nodes must arbitrate for access to the quorum disk resource. Quorum disk arbitration may occur within seconds. However, the performance of hardware and attached devices on the shared SCSI bus may influence the speed of this process.
All of these factors and many others may influence resource startup or failover time requirements. Hardware and network performance may also impact cluster resources.
Article ID: 171277 - Last Review: February 23, 2007 - Revision: 3.3