Article ID: 885688 - View products that this article applies to.
When you start your Microsoft Windows Server 2003-based computer that is running Windows Clustering, the following events may be logged.
Event ID: 57
Event Type: Warning
Event ID: 50
Event Type: Error
Event Source: NTFS
Event ID: 55
The file system structure on disk is corrupt and unusable. Please run the chkdsk utility on the volume "Drive_letter:"
Event Type: Warning
This behavior occurs when the cluster node computer starts. This behavior is caused by a conflict between the NTFS file system and the cluster services driver component (Clusdisk) filter.
This behavior does not indicate a problem with the cluster node hard disk and can be ignored on a correctly configured cluster.
However, if you have an incorrectly configured cluster, data corruption may occur. Use of the Microsoft Cluster Configuration Validation Wizard can provide a high level of confidence that your subsequent cluster installation and operation will be successful. The Microsoft Cluster Configuration Validation Wizard, "ClusPrep," is a validation tool that does a complete system inventory and runs focused tests on servers that are configured and ready for Microsoft Server Cluster installation (or, in other words, before the servers are a cluster). For more information about the tool, visit the following Microsoft Web site:
When the hard disk plug-and-play (PnP) arrival event occurs, the NTFS file system immediately tries to mount the file system. This operation creates several Input Output Control (IOCTL) function calls and several I/O procedures. Immediately after the NTFS file system begins the mounting sequence, the Clusdisk responds to the PnP arrival event and blocks all I/O operations to the disk subsystem. This blocks all IOCTL and I/O requests, including the pending requests. Because of this behavior, the volume manager disk driver (ftdisk) interprets the block as an I/O error and records the event ID 57, event ID 55, and event ID 50 entries.
Note The Clusdisk load delay is less than 10 milliseconds. In this short time, the NTFS file system cannot mount. Therefore, this conflict cannot pose a problem to the NTFS file system.
In this situation, the event is logged only when the following conditions are true:
Note When there are multiple nodes, the reservation on the disk normally prevents the other nodes from writing to the disk. However, if the nodes are unable to communicate through the cluster networks, then a "split-brain" scenario may occur. A node that is isolated from the other nodes when it is started will not be able to communicate with the owning node and will try to form the cluster itself, which may break the reservation on some or all cluster disks. For more information, click the following article number to view the article in the Microsoft Knowledge Base:
(http://support.microsoft.com/kb/309186/ )How the Cluster service reserves a disk and brings a disk online
Article ID: 885688 - Last Review: August 4, 2009 - Revision: 2.0