Question

I am trying to work on the upgrade component in our program.
This requires changing an index’s name

EXEC sp_rename N'Sig_Summary1Index.IX_Sig_Summary1Index_StartTime',

N'Sig_Summary3Index.IX_Sig_Summary1Index1_StartTime', N'INDEX';

(we need to remain compatible with SQL SERVER 2005)

and then check if it exists in the same transaction:

IF NOT EXISTS 
(   SELECT * 
    FROM sys.indexes 
    WHERE object_id = OBJECT_ID(N'[dbo].[Sig_Summary1Index]' ) 
        AND name = N'[IX_Sig_Summary1Index1_StartTime]') 

CREATE CLUSTERED INDEX [IX_Sig_Summary1Index1_StartTime] 
ON [dbo].[Sig_Summary1Index] (  
    [StartTime] ASC 
)
WITH (  
    PAD_INDEX  = OFF, 
    STATISTICS_NORECOMPUTE  = OFF, 
    SORT_IN_TEMPDB = OFF, 
    IGNORE_DUP_KEY = OFF, 
    DROP_EXISTING = OFF, 
    ONLINE = OFF, 
    ALLOW_ROW_LOCKS  = ON, 
    ALLOW_PAGE_LOCKS  = ON, 
    FILLFACTOR = 95
) 
ON [PRIMARY]

This WILL try to create the index. 🙁
somehow, during the transaction the sys.indexes table is not updated yet.
Is there any other way to avoid creating this index?

Edit: Is there any way i can “flush” the sys tables to update? Insert the records manually? do something so that the “check” will not be changed?

Answer 1

sys.tables is always in perfect synch with CREATE INDEX. If your query doesn’t find it means it looks for the wrong index name. If you run the CREATE INDEX fron inside a transaction and then, before commit, you look in sys.tables from another transaction, you fall into ordinary read isolation levels and you won’t see the new index until the transaction commits. Which is the expected and desired behavior.