Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

sql try/catch rollback/commit - preventing erroneous commit after rollback

People also ask

What happens if you rollback after commit?

After you commit the transaction, the changes are visible to other users' statements that execute after the commit. You can roll back (undo) any changes made during the transaction with the ROLLBACK statement (see ROLLBACK.

Can we execute commit immediately after rollback?

I have confirmed that after rollback we cannot commit the same transaction. Make sure another transaction is not in waiting, else it will be committed.

What happens when a rollback transaction is issued in a trigger?

If a ROLLBACK TRANSACTION is issued in a trigger: All data modifications made to that point in the current transaction are rolled back, including any made by the trigger. The trigger continues executing any remaining statements after the ROLLBACK statement.

Is it correct best practice to have the try catch block inside the transaction or should the transaction be inside the try block?

Option A is the correct choice. It is possible for all statements in a transaction to work and then the actual COMMIT to fail, so you keep the COMMIT inside your TRY block so that any failure of the COMMIT will be caught and you can gracefully handle this error and rollback.


I always thought this was one of the better articles on the subject. It includes the following example that I think makes it clear and includes the frequently overlooked @@trancount which is needed for reliable nested transactions

PRINT 'BEFORE TRY'
BEGIN TRY
    BEGIN TRAN
     PRINT 'First Statement in the TRY block'
     INSERT INTO dbo.Account(AccountId, Name , Balance) VALUES(1, 'Account1',  10000)
     UPDATE dbo.Account SET Balance = Balance + CAST('TEN THOUSAND' AS MONEY) WHERE AccountId = 1
     INSERT INTO dbo.Account(AccountId, Name , Balance) VALUES(2, 'Account2',  20000)
     PRINT 'Last Statement in the TRY block'
    COMMIT TRAN
END TRY
BEGIN CATCH
    PRINT 'In CATCH Block'
    IF(@@TRANCOUNT > 0)
        ROLLBACK TRAN;

    THROW; -- raise error to the client
END CATCH
PRINT 'After END CATCH'
SELECT * FROM dbo.Account WITH(NOLOCK)
GO

In your first example, you are correct. The batch will hit the commit transaction, regardless of whether the try block fires.

In your second example, I agree with other commenters. Using the success flag is unnecessary.

I consider the following approach to be, essentially, a light weight best practice approach.

If you want to see how it handles an exception, change the value on the second insert from 255 to 256.

CREATE TABLE #TEMP ( ID TINYINT NOT NULL );
INSERT  INTO #TEMP( ID ) VALUES  ( 1 )

BEGIN TRY
    BEGIN TRANSACTION

    INSERT  INTO #TEMP( ID ) VALUES  ( 2 )
    INSERT  INTO #TEMP( ID ) VALUES  ( 255 )

    COMMIT TRANSACTION
END TRY
BEGIN CATCH
    DECLARE 
        @ErrorMessage NVARCHAR(4000),
        @ErrorSeverity INT,
        @ErrorState INT;
    SELECT 
        @ErrorMessage = ERROR_MESSAGE(),
        @ErrorSeverity = ERROR_SEVERITY(),
        @ErrorState = ERROR_STATE();
    RAISERROR (
        @ErrorMessage,
        @ErrorSeverity,
        @ErrorState    
        );
    ROLLBACK TRANSACTION
END CATCH

SET NOCOUNT ON

SELECT ID
FROM #TEMP

DROP TABLE #TEMP

I used below ms sql script pattern several times successfully which uses Try-Catch,Commit Transaction- Rollback Transaction,Error Tracking.

Your TRY block will be as follows

 BEGIN TRY
 BEGIN TRANSACTION T
 ----
 //your script block
 ----
 COMMIT TRANSACTION T 
 END TRY

Your CATCH block will be as follows

BEGIN CATCH
DECLARE @ErrMsg NVarChar(4000), 
        @ErrNum Int, 
        @ErrSeverity Int, 
        @ErrState Int, 
        @ErrLine Int, 
        @ErrProc NVarChar(200)
 SELECT @ErrNum = Error_Number(), 
       @ErrSeverity = Error_Severity(), 
       @ErrState = Error_State(), 
       @ErrLine = Error_Line(), 
       @ErrProc = IsNull(Error_Procedure(), '-')
 SET @ErrMsg = N'ErrLine: ' + rtrim(@ErrLine) + ', proc: ' + RTRIM(@ErrProc) + ', 
       Message: '+ Error_Message()

Your ROLLBACK script will be part of CATCH block as follows

IF (@@TRANCOUNT) > 0 
BEGIN
PRINT 'ROLLBACK: ' + SUBSTRING(@ErrMsg,1,4000)
ROLLBACK TRANSACTION T
END
ELSE
BEGIN
PRINT SUBSTRING(@ErrMsg,1,4000);   
END

END CATCH

Above different script blocks you need to use as one block. If any error happens in the TRY block it will go the the CATCH block. There it is setting various details about the error number,error severity,error line ..etc. At last all these details will get append to @ErrMsg parameter. Then it will check for the count of transaction (@@TRANCOUNT >0) , ie if anything is there in the transaction for rollback. If it is there then show the error message and ROLLBACK TRANSACTION. Otherwise simply print the error message.

We have kept our COMMIT TRANSACTION T script towards the last line of TRY block in order to make sure that it should commit the transaction(final change in the database) only after all the code in the TRY block has run successfully.