What is the best practice use of SQL Server T-SQL error handling? [closed]

Question

We have a large application mainly written in SQL Server 7.0, where all database calls are to stored procedures. We are now running SQL Server 2005, which offers more T-SQL features.

After just about every SELECT, INSERT, UPDATE, and DELETE, the @@ROWCOUNT and @@ERROR get captured into local variables and evaluated for problems. If there is a problem the following is done:

• error message output parameter is set
• rollback (if necessary) is done
• info is written (INSERT) to log table
• return with a error number, unique to this procedure (positive if fatal, negative is warning)

They all don't check the rows (only when it is known) and some differ with more or less log/debug info. Also, the rows logic is somethimes split from the error logic (on updates where a concurrency field is checked in the WHERE clause, rows=0 means someone else has updated the data). However, here is a fairly generic example:

SELECT, INSERT, UPDATE, or DELETE  SELECT @Error=@@ERROR, @Rows=@@ROWCOUNT IF @Rows!=1 OR @Error!=0 BEGIN     SET @ErrorMsg='ERROR 20, ' + ISNULL(OBJECT_NAME(@@PROCID), 'unknown')                                 + ' - unable to ???????? the ????.'     IF @@TRANCOUNT >0     BEGIN          ROLLBACK     END      SET @LogInfo=ISNULL(@LogInfo,'')+'; '+ISNULL(@ErrorMsg,'')+         + ' @YYYYY='        +dbo.FormatString(@YYYYY)         +', @XXXXX='        +dbo.FormatString(@XXXXX)         +', Error='         +dbo.FormatString(@Error)         +', Rows='          +dbo.FormatString(@Rows)      INSERT INTO MyLogTable (...,Message) VALUES (....,@LogInfo)      RETURN 20  END 

I am looking into replacing how we do this with the TRY-CATCH T-SQL. I've read about the TRY...CATCH (Transact-SQL) syntax, so don't just post some summary of that. I'm looking for any good ideas and how best to do or improve our error handling methods. It doesn't have to be Try-Catch, just any good or best practice use of T-SQL error handling.

Answer 1

You should read this:

http://www.sommarskog.se/error-handling-I.html

I can't recommend that link highly enough. It's a bit long, but in a good way.

There's a disclaimer at the front that it was originally written for SQL Server 2000, but it covers the new try/catch error handling abilities in SQL Server 2005+ as well.

Answer 2

We currently use this template for any queries that we execute (you could leave out the Transaction stuff, if you don't need it in e.g. a DDL statement):

BEGIN TRANSACTION BEGIN TRY     // do your SQL statements here      COMMIT TRANSACTION END TRY BEGIN CATCH     SELECT          ERROR_NUMBER() AS ErrorNumber,         ERROR_SEVERITY() AS ErrorSeverity,         ERROR_STATE() AS ErrorState,         ERROR_PROCEDURE() AS ErrorProcedure,         ERROR_LINE() AS ErrorLine,         ERROR_MESSAGE() AS ErrorMessage      ROLLBACK TRANSACTION END CATCH 

Of course, you could easily insert the caught exception into your error log table.

It works really well for us. You could probably even automate some of the conversion from your old stored procs to a new format using Code Generation (e.g. CodeSmith) or some custom C# code.