Question

I recently had to switch from using Microsofts SQL server to using MySQL.

The application I wrote is using the Entity Framework 4.0. I really like MySQL but the process has been a bit of a pain. Most of the CRUD operations are done with stored procedures so I had fun in my encounter with MySQL bug 55778.

I got over that and manually edited the EDMX file.

Everything worked well until in one scenario I started getting an optimistic concurrency exception.

This is where is happened:

using (var context = new DBEntities())
            {
                context.Requests.AddObject(request);
                context.SaveChanges();

                response.RequestId = request.Id;
                context.Responses.AddObject(response);
                context.SaveChanges();

            } 

The request is logged fine. The exception is thrown when I try to save changes on the response. RequestId is the primary key (not auto incremented) of the response table. I tried to use MySQL profiler to see what was happening but it didn’t really work. I could only profile queries made through my terminal but not queries made by the application running on my local IIS.

I ended up using Wireshark. I found that when I went to save the response the stored procedure linked to Insert on the response Entity was being called. MySQL returned a response saying 0 rows were effected. This caused the exception. The application then called ROLLBACK removing any changes that should have been made.

If I run that stored procedure in a terminal with the exact same values it works.

So over the wire I see this being called:

CALL `DevDB`.`LogResponse` (1,'2012-03-06 12:30',1,1,'test','test','test','test',false,'test');

Then 0 rows effected being returned and a ROLLBACK being called.

Running the very same command from MySQL workbench works. I have tried this multiple times, dropping the database and starting over. I can’t find a reason for this. The request is logged using a similar method and it works.

The application has EXECUTE permissions on both stored routines.

Has anyone come across something like this before?

EDIT:
I also use MySQL handlers to catch any error I could think of and log it to an error log table.
e.g.

DECLARE EXIT HANDLER FOR SQLEXCEPTION
BEGIN
    ROLLBACK;
    INSERT INTO `DevDB`.`tblDatabaseErrorLogs`(ERROR_MESSAGE, ERROR_CODE)
    VALUES(
    'SQL Exception at LogResponse',    
    'Unknown'
    );
END; 

So far none to the error codes I planned for – SQL Exception, SQL WARNING and a bunch of others – appear to have happened as the error log table is empty.

Answer 1

I discovered what was causing the problem but I don’t really understand why. If someone gives a good explanation then I will accept that as the answer to this question rather than accepting my own.

This is what appears to have caused the problem.

When using stored procedures to do inserts in the Entity Framework I found that to give the Entity in the application the correct Id at run time I had to return the Id in the stored procedure.

In MySQL I do it like this:

CREATE PROCEDURE LogResponse(IN myVals)
BEGIN 
Insert Into `DevDB`.`tblResponses`
(
    someValueNames
)
Values
(
    someValues
);

 SELECT LAST_INSERT_ID() As NewResponseId;
END

The returned SELECT value maps the the PK Id of the Entity which uses this stored procedure as its INSERT function.

It just happened that I didn’t use the return value for the response because I didn’t need the Id in the application. This was an oversight I guess but it didn’t raise any issue when I did a similar thing with SQL Server

i.e.

SELECT SCOPE_IDENTITY() As NewResponseId

All I did to fix the thing is remove the line

SELECT LAST_INSERT_ID() As NewResponseId;

Everything appears to be working as it should now.