Sql-server – Incorrect query results for execution plan, unsure of exact cause of problem

sql serversql-server-2012

We had a query yesterday that was intended to page through results but on the first page of results the query only returned 4 records instead of the expect 25. Changing the query in any way resulted in 25 records which meant to me that it was an execution plan issue. I'm not familiar with a way to view execution plan in production and running the query in sql studio didn't result in same problem, probably due to minor differences.

I read suggestions that such a thing could be caused by a corrupted index. I ran a checkdb on the database and it found no errors. In the end I cleared the execution plans and all has been well.

If it wasn't a corruption issue of some kind and only a problem for the very specific execution plan then does that mean there was simply an error in the execution plan and we are hitting a bug in SQL Server? We are on SQL Server 2012 RTM with no updates so I looked through documentation on all fixes in cumulative updates and service packs but none of the issues appeared to related to our own.

Any other ideas or thoughts as to what could cause this?

(@P1 varchar(8000),@P2 bit,@P3 varchar(8000),@P4 bit,@P5 varchar(8000))

SELECT e.*
FROM (

    SELECT 

        TOP 25
        ROW_NUMBER() OVER (
            ORDER BY AddedDate DESC
        ) AS Row,

        ID
        , Prefix
        , FirstName
        , LastName
        , Company
        , Address
        , City
        , State
        , Zip
        , Country
        , WorkPhone
        , HomePhone
        , MobilePhone
        , Email
        , MailingLists
        , AddedDate
        , AwaitingOptin
        , OptInDate
        , Processed
        , ProcessedDate
        , Deleted
        , Source
        , GRRecID
        , DatabaseID
        , (SELECT COUNT(*) 
            FROM TableA
            LEFT OUTER JOIN TableB ON TableA.GRRecID = TableB.GRRecID
            WHERE TableA.AccountID = @P1 
                AND TableA.Email = TableC.Email
                AND TableA.Deleted = @P2 
                ) AS Matches
    FROM TableC
    WHERE AccountID = @P3  
        AND Processed = @P4  
        AND Deleted = 0
        AND Source = @P5  
) AS e
WHERE Row >= 1 AND Row <= 25
ORDER BY Row

Best Answer

You should consider refactoring to use the new paging clause in SQL Server 2012, ( OFFSET...FETCH ), rather than that old-fashioned 'roll your own paging' eg

SELECT 
...
FROM TableC
WHERE AccountID = @P3
    AND Processed = @P4
    AND Deleted = 0
    AND Source = @P5  

ORDER BY ProcessedDate DESC
OFFSET 0 ROWS FETCH NEXT 25 ROWS ONLY

It's potentially more efficient. Here's a sample SQL Fiddle.

Related Solutions

Sql-server – Should an INSERT cause an exclusive lock on a foreign key

Here's my first recommendation to remove the deadlock (and this is a very common cause with the same resolution every time). Your RoutePlan_Save procedure currently has this logic:

IF EXISTS (SELECT ... WHERE key = @key)
  UPDATE ... WHERE key = @key;
ELSE
  INSERT(key) VALUES(@key);

So you actually have two queries here that require locks on the table, regardless of whether you are ultimately going to perform an insert or update. Stop doing that! What is the point of checking if a row exists, and then in a separate query, firing an update statement that again has to check if the row exists? This requires more resources, not less, than just trying to update the row not knowing whether it exists. So do this instead:

BEGIN TRANSACTION;
UPDATE ... WHERE key = @key;
IF @@ROWCOUNT = 0
BEGIN
  INSERT(key) VALUES(@key);
END
COMMIT TRANSACTION;

You can also isolate the other read-only processes by using a different isolation level that doesn't require them to wait for write operations to complete. NOLOCK / READ UNCOMMITTED is a popular one, but READ COMMITTED SNAPSHOT - assuming you can take the hit on tempdb - is way better IMHO. There are other alternatives, too. I happened to blog about this just yesterday:

Bad habits : Putting NOLOCK everywhere

Sql-server – Is execution plan cached “better” for stored procedures than for a non-dynamic query

_{The answer is also available as a standalone blog article.}

In order to find it out, I did some tests. The goal is to have the same parametrized query executed either directly from C# or by calling a stored procedure and to compare the runtime performance.

I started to create a stored procedure which does a sample query using Adventure Works database:

create procedure Demo
    @minPrice int 
as
begin
    set nocount on;

    select top 1 [p].[Name], [p].[ProductNumber], [ph].[ListPrice]
    from [Production].[Product] p
    inner join [Production].[ProductListPriceHistory] ph
    on [p].[ProductID] = ph.[ProductID]
    and ph.[StartDate] =
    (
        select top 1 [ph2].[StartDate]
        from [Production].[ProductListPriceHistory] ph2
        where [ph2].[ProductID] = [p].[ProductID]
        order by [ph2].[StartDate] desc
    )
    where [p].[ListPrice] > @minPrice
end

Then, I use the following piece of code to compare the performances:

long RunQuery(SqlConnection connection, int minPrice)
{
    const string Query = @"
    select top 1 [p].[Name], [p].[ProductNumber], [ph].[ListPrice]
    from [Production].[Product] p
    inner join [Production].[ProductListPriceHistory] ph
    on [p].[ProductID] = ph.[ProductID]
    and ph.[StartDate] =
    (
        select top 1 [ph2].[StartDate]
        from [Production].[ProductListPriceHistory] ph2
        where [ph2].[ProductID] = [p].[ProductID]
        order by [ph2].[StartDate] desc
    )
    where [p].[ListPrice] > @minPrice
    option (recompile)";

    using (var command = new SqlCommand(Query, connection))
    {
        command.Parameters.AddWithValue("@minPrice", minPrice);
        var stopwatch = Stopwatch.StartNew();
        command.ExecuteNonQuery();
        stopwatch.Stop();
        return stopwatch.ElapsedMilliseconds;
    }
}

long RunStoredProcedure(SqlConnection connection, int minPrice)
{
    using (var command = new SqlCommand("exec Demo @minPrice with recompile", connection))
    {
        command.Parameters.AddWithValue("@minPrice", minPrice);
        var stopwatch = Stopwatch.StartNew();
        command.ExecuteNonQuery();
        stopwatch.Stop();
        return stopwatch.ElapsedMilliseconds;
    }
}

ICollection<long> Execute(Func<SqlConnection, int, long> action)
{
    using (var connection = new SqlConnection("Server=.;Database=AdventureWorks2014;Trusted_Connection=True;"))
    {
        connection.Open();
        using (var command = new SqlCommand("DBCC FreeProcCache; DBCC DropCleanbuffers;", connection))
        {
            command.ExecuteNonQuery();
        }

        return Enumerable.Range(0, 100).Select(i => action(connection, i)).ToList();
    }
}

void Main()
{
    var queries = Execute(RunQuery);
    var storedProcedures = Execute(RunStoredProcedure);

    Console.WriteLine("Stored procedures: {0} ms. Details: {1}.", storedProcedures.Sum(), string.Join(", ", storedProcedures));
    Console.WriteLine("Queries: {0} ms. Details: {1}.", queries.Sum(), string.Join(", ", queries));
}

Notice option (recompile) and with recompile. This will force SQL Server to discard previously cached execution plans.

Each query is run une hundred times with a different parameter every time. The time spent by the server is measured at client side.

By running DBCC FreeProcCache; DBCC DropCleanbuffers; before gathering metrics, I make sure that all previously cached execution plans are removed.

Running this code gives the following output:

Stored procedures: 786 ms. Details: 12, 7, 7, 9, 7, 7, 9, 8, 8, 6, 8, 9, 8, 8, 14, 8, 7, 8, 7, 10, 10, 7, 9, 6, 9, 8, 8, 7, 7, 10, 8, 7, 7, 6, 7, 8, 8, 7, 7, 7, 14, 8, 8, 8, 7, 9, 8, 8, 7, 6, 6, 12, 7, 7, 8, 7, 8, 7, 8, 6, 7, 7, 7, 12, 8, 6, 6, 7, 8, 7, 8, 8, 7, 11, 8, 7, 8, 8, 7, 9, 8, 9, 10, 8, 7, 7, 8, 8, 7, 9, 7, 6, 9, 7, 6, 9, 8, 6, 6, 6.
Queries: 799 ms. Details: 21, 8, 8, 7, 6, 6, 11, 7, 6, 6, 9, 8, 8, 7, 9, 8, 7, 7, 7, 7, 7, 7, 10, 8, 8, 7, 8, 7, 6, 11, 19, 10, 8, 7, 8, 7, 7, 7, 6, 9, 7, 9, 7, 7, 8, 7, 12, 9, 7, 7, 7, 8, 7, 7, 8, 7, 7, 7, 9, 8, 7, 7, 7, 6, 7, 7, 16, 7, 7, 7, 8, 8, 9, 8, 7, 9, 8, 7, 8, 7, 7, 6, 7, 7, 7, 7, 12, 7, 9, 9, 7, 7, 7, 7, 9, 8, 7, 8, 11, 8.

Let's run it again:

Stored procedures: 763 ms. Details: 11, 8, 10, 8, 8, 14, 10, 6, 7, 7, 6, 7, 7, 9, 6, 6, 6, 8, 6, 6, 7, 6, 8, 7, 16, 8, 7, 8, 9, 7, 7, 8, 7, 7, 11, 10, 7, 6, 7, 8, 7, 7, 7, 7, 7, 7, 10, 9, 9, 7, 6, 7, 6, 7, 7, 6, 6, 6, 6, 6, 10, 9, 10, 7, 6, 6, 6, 6, 6, 8, 7, 6, 6, 7, 8, 9, 7, 8, 7, 10, 7, 7, 7, 6, 7, 6, 7, 11, 13, 8, 7, 10, 9, 8, 8, 7, 8, 7, 7, 7.
Queries: 752 ms. Details: 25, 10, 8, 8, 12, 8, 7, 9, 9, 8, 6, 7, 7, 6, 8, 6, 7, 7, 8, 9, 7, 7, 7, 7, 6, 10, 8, 7, 7, 7, 7, 7, 7, 7, 8, 9, 7, 6, 6, 6, 7, 13, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 6, 10, 7, 7, 8, 9, 8, 7, 6, 6, 7, 7, 9, 7, 8, 6, 9, 7, 7, 8, 7, 6, 6, 7, 7, 7, 7, 6, 7, 7, 8, 7, 7, 6, 7, 9, 8, 7, 7, 7, 7, 6, 7, 6, 6, 9, 7, 7.

It seems that the performance is very close between stored procedures and direct queries. Running the code a dozen times, I notice that stored procedures seem to be slightly fast, but the gap is very narrow. Possibly passing around the whole query creates this additional cost, which may increase if SQL Server is hosted on a dedicated machine with a slow LAN between it and the application server.

Let's now turn execution plan caching on and see what happens. To do this, I remove option (recompile) and with recompile from the code. Here's the new output:

Stored procedures: 26 ms. Details: 23, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.
Queries: 15 ms. Details: 14, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.

It becomes clear that caching has exactly the same effect for both direct queries and stored procedures. In both cases, it reduces the time to nearly zero milliseconds, and the most expensive query is the first one—the one which runs after the removal of cached execution plans.

Running the same code again shows a similar pattern. Sometimes, queries are faster, and sometimes stored procedures are. But every time, the first query is the most expensive one, and all others are close to zero milliseconds.

Reopening SQL connection

If the SQL connection is opened for every query, such as in this slightly modified code:

long RunQuery(string connectionString, int minPrice)
{
    const string Query = @"
    select top 1 [p].[Name], [p].[ProductNumber], [ph].[ListPrice]
    from [Production].[Product] p
    inner join [Production].[ProductListPriceHistory] ph
    on [p].[ProductID] = ph.[ProductID]
    and ph.[StartDate] =
    (
        select top 1 [ph2].[StartDate]
        from [Production].[ProductListPriceHistory] ph2
        where [ph2].[ProductID] = [p].[ProductID]
        order by [ph2].[StartDate] desc
    )
    where [p].[ListPrice] > @minPrice
    option (recompile)";

    using (var connection = new SqlConnection(connectionString))
    {
        connection.Open();
        using (var command = new SqlCommand(Query, connection))
        {
            command.Parameters.AddWithValue("@minPrice", minPrice);
            var stopwatch = Stopwatch.StartNew();
            command.ExecuteNonQuery();
            stopwatch.Stop();
            return stopwatch.ElapsedMilliseconds;
        }
    }
}

long RunStoredProcedure(string connectionString, int minPrice)
{
    using (var connection = new SqlConnection(connectionString))
    {
        connection.Open();
        using (var command = new SqlCommand("exec Demo @minPrice with recompile", connection))
        {
            command.Parameters.AddWithValue("@minPrice", minPrice);
            var stopwatch = Stopwatch.StartNew();
            command.ExecuteNonQuery();
            stopwatch.Stop();
            return stopwatch.ElapsedMilliseconds;
        }
    }
}

ICollection<long> Execute(Func<string, int, long> action)
{
    var connectionString = "Server=.;Database=AdventureWorks2014;Trusted_Connection=True;";
    using (var connection = new SqlConnection(connectionString))
    {
        connection.Open();
        using (var command = new SqlCommand("DBCC FreeProcCache; DBCC DropCleanbuffers;", connection))
        {
            command.ExecuteNonQuery();
        }
    }

    return Enumerable.Range(0, 100).Select(i => action(connectionString, i)).ToList();
}

void Main()
{
    var queries = Execute(RunQuery);
    var storedProcedures = Execute(RunStoredProcedure);

    Console.WriteLine("Stored procedures: {0} ms. Details: {1}.", storedProcedures.Sum(), string.Join(", ", storedProcedures));
    Console.WriteLine("Queries: {0} ms. Details: {1}.", queries.Sum(), string.Join(", ", queries));
}

the observed metrics are very similar:

Stored procedures: 748 ms. Details: 11, 8, 6, 6, 8, 9, 9, 8, 8, 7, 6, 8, 7, 9, 6, 6, 6, 6, 6, 6, 7, 7, 6, 9, 6, 6, 7, 6, 6, 7, 8, 6, 7, 7, 7, 13, 7, 7, 8, 7, 8, 8, 7, 7, 7, 7, 6, 7, 8, 8, 8, 9, 7, 6, 8, 7, 6, 7, 6, 6, 6, 6, 8, 12, 7, 9, 9, 6, 7, 7, 7, 8, 10, 12, 8, 7, 6, 9, 8, 7, 6, 6, 7, 8, 6, 6, 12, 7, 8, 10, 10, 7, 8, 7, 8, 10, 8, 7, 8, 7.
Queries: 761 ms. Details: 31, 9, 7, 6, 6, 8, 7, 7, 7, 7, 7, 6, 8, 7, 6, 6, 7, 10, 8, 10, 9, 7, 7, 7, 7, 10, 13, 7, 10, 7, 6, 6, 6, 8, 7, 7, 7, 7, 7, 7, 7, 9, 7, 7, 7, 6, 6, 6, 9, 7, 7, 7, 7, 7, 6, 8, 10, 7, 7, 7, 7, 7, 7, 7, 8, 6, 10, 10, 7, 8, 8, 7, 7, 7, 7, 7, 6, 6, 7, 6, 8, 7, 7, 7, 7, 7, 7, 7, 8, 7, 8, 7, 9, 7, 6, 6, 12, 10, 7, 6.

with option (recompile) and with recompile and:

Stored procedures: 15 ms. Details: 14, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.
Queries: 32 ms. Details: 26, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0.

without.

Under the hood

Let's see what happens under the hood. The following query shows cached execution plans:

select usecounts, size_in_bytes, cacheobjtype, objtype, text 
from sys.dm_exec_cached_plans 
cross apply sys.dm_exec_sql_text(plan_handle)
where cacheobjtype = 'Compiled Plan'
order by usecounts desc

When running this query after executing the stored procedures one hundred times, the result of the query looks like this:

usecounts   size_in_bytes cacheobjtype                                       objtype              text
----------- ------------- -------------------------------------------------- -------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
100         90112         Compiled Plan                                      Proc                 create procedure Demo
    @minPrice int 
as
begin
    set nocount on;

    select top 1 [p].[Name], [p].[ProductNumber], [ph].[ListPrice]
    from [Production].[Product] p
    inner join [Production].[ProductListPriceHistory] ph
    on [p].[ProductID] = ph.[Product
100         16384         Compiled Plan                                      Prepared             (@minPrice int)exec Demo @minPrice --with recompile
1           49152         Compiled Plan                                      Adhoc                --DBCC FreeProcCache
--DBCC DropCleanbuffers

select usecounts, size_in_bytes, cacheobjtype, objtype, text 
from sys.dm_exec_cached_plans 
cross apply sys.dm_exec_sql_text(plan_handle)
where cacheobjtype = 'Compiled Plan'
order by usecounts desc

(3 row(s) affected)

When running the query directly one hundred times, the result is:

usecounts   size_in_bytes cacheobjtype                                       objtype              text
----------- ------------- -------------------------------------------------- -------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
100         73728         Compiled Plan                                      Prepared             (@minPrice int)
    select top 1 [p].[Name], [p].[ProductNumber], [ph].[ListPrice]
    from [Production].[Product] p
    inner join [Production].[ProductListPriceHistory] ph
    on [p].[ProductID] = ph.[ProductID]
    and ph.[StartDate] =
    (
        select top 1 [ph2].[
1           49152         Compiled Plan                                      Adhoc                --DBCC FreeProcCache
--DBCC DropCleanbuffers

select usecounts, size_in_bytes, cacheobjtype, objtype, text 
from sys.dm_exec_cached_plans 
cross apply sys.dm_exec_sql_text(plan_handle)
where cacheobjtype = 'Compiled Plan'
order by usecounts desc

(2 row(s) affected)

Conclusion

The execution plan is cached for stored procedures and direct queries.
The performance between stored procedures and direct queries is very similar when the SQL Server and the application are hosted on the same machine. When SQL Server is hosted on a dedicated server accessed through LAN, using stored procedures may result in better performance.