Will creating a nonclustered index break change tracking on my database (SQL Server 2008)?
I have been investigating some performance issues with my database (SQL Server 2008). SQL Management studio suggested that I use this code to add a nonclustered index which will decrease processing time by over 90%.
USE [DatabaseName]
GO
CREATE NONCLUSTERED INDEX [<Name of Missing Index, sysname,>]
ON [dbo].[tblAnswers] ([Quest开发者_运维技巧ionID])
INCLUDE ([CallID],[Value])
GO
I am however worried that using this code will break change tracking on my database which is essential to keep in tact.
Having tested the code an another database it seemed okay, the resulting consequence was that the next time i synchronized (from my occasionally connected client) it took much longer to process despite not actually uploading or downloading any extra data, subsequent synchronizations returned to usual speed.
Is this code safe to run?
For anyone interested, i went ahead after some more testing and bit the bullet.
I set SQL Management studio to inform me of any action that would cause tables to be dropped / recreated. I then ran the code and it was fine. No tables were affected in this way and so : -
change tracking remained in tact.
Phew.
精彩评论