Can anyone please tell me why does the following statement inside a given stored procedure returns repeated results even with locks on the rows used by the first SELECT statement?
BEGIN TRANSACTION
DECLARE @Temp TABLE ( ID INT )
INSERT INTO @Temp SELECT ID FROM SomeTable WITH (ROWLOCK, UPDLOCK, READPAST) WHERE SomeValue <= 10
INSERT INTO @Temp SELECT ID FROM SomeTable WITH (ROWLOCK, UPDLOCK, READPAST) WHERE SomeValue >= 5
SELECT * FROM @Temp
COMMIT TRANSACTION
Any values in SomeTable for which SomeValue is between 5 and 10 will be returned twice, even though they were locked in the first SELECT. I thought that locks were in place for the whole transaction, and so I wasn’t expecting the query to return repeated results. Why is this happening?
In the same session/transaction, READPAST won’t have any effect.
So even though you have ROWLOCK and UPDLOCK for the first SELECT , the next SELECT ignores these. If the 2nd SELECT was a different session, then it’d work as expected.
See these for more info: ORDER BY and WITH(ROWLOCK, UPDLOCK, READPAST) and TABLOCKX versus SERIALIZABLE