Microsoft has released a hotfix for an index-scan error in UPDATE operations. When you run an UPDATE statement and the connection's isolation level is set to READ UNCOMMITTED, you might receive the error message "Error: 644, Severity: 21, State: 6. Could not find the index entry for RID '%.*hs' in index page % S_PGID, index ID %d, database '%.*ls'." This problem only occurs when you use an index scan as part of the UPDATE operation and perform the index scan as a parallel operation. When the isolation level is set to READ UNCOMMITTED, locks aren't held on rows that qualify for the scan from the time of the scan to the update. This behavior lets other connections modify a row that's already qualified for the update before the UPDATE operation is complete. In this situation, you can safely ignore the 644 error message. The UPDATE statement correctly reports that a specific row can't be found and terminates--any changes that might have been made are rolled back. The hotfix now available from Microsoft ensures that the locks are held on qualifying rows from the time of the scan to the update. To learn more about the problem and the supported hotfix, read the Microsoft article "FIX: You receive a 644 error message when you run an UPDATE statement and the isolation level is set to READ UNCOMMITTED" at<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

   http://support.microsoft.com/?kbid=834290