Cluster Service May Stop After Failover

• Microsoft Windows NT Server, Enterprise Edition version 4.0 - Microsoft Cluster Server

SYMPTOMS

If you install Microsoft Cluster Server and shut down one of the nodes in the cluster, the cluster service may fail with the following event codes in the system event log:

• Event 1006
• Event 7023
• Event 1009

CAUSE

When you shut down one of the cluster nodes, all resources should transition to a surviving node in the cluster. If the shared SCSI bus loses termination, the surviving node may not have access to the shared devices. As a result, the surviving node may also fail and remove itself from the cluster -- only if the quorum disk is inaccessible.

The cluster service requires access to a shared SCSI bus. The bus must be terminated at all times. Some SCSI host adapters may provide termination only while the computer has power. Thus, after powering off, termination may be lost.

RESOLUTION

To alleviate this condition, supply external terminators for the bus or enable constant termination of the adapter.

MORE INFORMATION

Incorrect cabling or termination of the shared SCSI bus may also trigger this condition. Additionally, use of standard SCSI terminators with a Wide SCSI bus may provide inadequate termination. Some older terminators do not terminate the high bits on the Wide SCSI chain and may result in partial termination of the SCSI bus. Consult your hardware vendor for proper cabling and termination.

For additional information, please see the following article(s) in the Microsoft Knowledge Base:

   ARTICLE-ID: Q168094
   TITLE     : Problems Attaching SCSI-2 Devices to SCSI-3 Adapter