Single user mode "missing behavior" on SQL Server

Question

I run a script provided by one of Microsoft employee to find out about which indexes need to Rebuild/Reorganize depending on the average fragmentation. I got back a reasonable list but while trying to rebuild some of them on a specific database I kept receiving errors :

The first idea I got is to set the database in single user mode, rebuild the indexes and then bring it back to life. Well that did not help because the database is being populated by a Windows service that ironically uses the same user I am connected with and the only available to me with enough permissions to do so. I am working on a corporate environment so the moon is a bit closer than getting another user credentials. I also cannot stop the service while executing my tasks because it is used for many other things.

My question is simple: How can I force single-user mode to force single connection source? In other words how to hide the database or eventually the SQL server from the service? It will correctly handle the absence as a network issue so I don't have to worry about that part.

Answer 1

I found a good solution to use that might help others. I start by getting the list of transactions with locks on the current table using :

USE [Your DB Name]

SELECT REQUEST_MODE, REQUEST_TYPE, REQUEST_SESSION_ID
FROM sys.dm_tran_locks
WHERE RESOURCE_TYPE = 'OBJECT'
AND RESOURCE_ASSOCIATED_ENTITY_ID =(SELECT OBJECT_ID('YourTableName'))

The REQUEST_SESSION_ID is the ID of the session which has the lock set on the table. Then I run EXEC sp_who2 to make sure that the SPID is the one for the expected service. All I needed to do at the end was KILL <SPID> and rebuild the index. You might need to do it multiple time if you are building more than one index as the lock could be set again.

Answer 2

There is an ONLINE = ON/OFF option available when rebuilding indexes in SQL Server 2005 and above which controls how users can access underlying table which may solve your problem.

http://msdn.microsoft.com/en-us/library/ms188388(v=sql.110).aspx

Answer 3

your problem is that the interface will only wait a certain amount of time before deciding to fail. I run into this all of the time.

You can try scripting the change and then running it manually, this will allow you to just wait until all of the locks are released by the users currently using the index. You will have to be careful though, an index rebuild locks the index for the time that it is running (unless of course you have enterprise edition, where rebuilds are online, and everything is made of money)