Clustering Cannot Determine If a Shared Disk Is Working Properly

MORE INFORMATION

The cluster service, which implements IsAlive poll operation, does not actually test the volumes on a cluster disk resource for file readability. File access is not checked because no volume on a disk resource is required to have files on it; if a disk has no files, the check for file readability would report that as a failure and try to fix the problem.

Instead, the cluster service checks to determine if the root directory of each volume is accessible. If no problem is found with this check the volume remains online. Therefore, IsAlive may return a SUCCESS value on a bad drive until an actual input/output (I/O) operation does not succeed. If a problem is found with the root directory check, a failure notification is sent to the cluster service, which takes the resource offline and then attempt to bring it online again. The cluster service checks to see if the dirty bit is set and if found, attempts to run Chkdsk.exe with the /f switch on the volume to fix it.

The only exception to this behavior is the volume that contains the quorum log. When the disk resource that contains this volume is brought online the files in the MSCS folder (or the folder specified in the quorum resource properties) are checked.

If Chkdsk.exe cannot be run on the volume, the cluster logging function adds the following entry to the cluster log:

   FixCorruption: CHKDSK returned status of <error code>
				

where an error code of zero indicates success.