SQL Server - stored procedure suddenly become slow

Question

I have written a stored procedure that, yesterday, typically completed in under a second. Today, it takes about 18 seconds. I ran into the problem yesterday as well, and it seemed to be solved by DROPing and re-CREATEing the stored procedure. Today, that trick doesn't appear to be working. :(

Interestingly, if I copy the body of the stored procedure and execute it as a straightforward query it completes quickly. It seems to be the fact that it's a stored procedure that's slowing it down...!

Does anyone know what the problem might be? I've searched for answers, but often they recommend running it through Query Analyser, but I don't have have it - I'm using SQL Server 2008 Express for now.

The stored procedure is as follows;

ALTER PROCEDURE [dbo].[spGetPOIs]
    @lat1 float,
    @lon1 float,
    @lat2 float,
    @lon2 float,
    @minLOD tinyint, 
    @maxLOD tinyint,
    @exact bit
AS
BEGIN
    -- Create the query rectangle as a polygon
    DECLARE @bounds geography;
    SET @bounds = dbo.fnGetRectangleGeographyFromLatLons(@lat1, @lon1, @lat2, @lon2);

    -- Perform the selection
    if (@exact = 0)
    BEGIN
        SELECT [ID], [Name], [Type], [Data], [MinLOD], [MaxLOD], [Location].[Lat] AS [Latitude], [Location].[Long] AS [Longitude], [SourceID]
        FROM [POIs]
        WHERE
            NOT ((@maxLOD  [MaxLOD])) AND
            (@bounds.Filter([Location]) = 1)
    END
    ELSE
    BEGIN
        SELECT [ID], [Name], [Type], [Data], [MinLOD], [MaxLOD], [Location].[Lat] AS [Latitude], [Location].[Long] AS [Longitude], [SourceID]
        FROM [POIs]
        WHERE
            NOT ((@maxLOD  [MaxLOD])) AND
            (@bounds.STIntersects([Location]) = 1)
    END

END

The 'POI' table has an index on MinLOD, MaxLOD, and a spatial index on Location.

Lactation

Answer 1

Ah, can it be the query plan sucks?

SP's get compiled / query plan determined on FIRST USE - depending on parameters. So, the parameters of the first call (when no plan is present) determine the query plan. At one point I gets dropped from cache, new plan generated.

Next time it runs slow, possibly make a call using query analyzer and get the selected plan - and check how it looks.

if it is this - put in an option to recompile the SP on every call (with recompile).

Seavey

This may well be it. I altered the stored procedure to recompile with every call and it seems (FOR NOW!) to have sped things up significantly. Thanks! Is there a way to avoid recompilation AND bad query plans? – Lactation

No. What you can do is dump the stored procedure and use dynamic sql... so the query plans get regenerated every call. This is a known issue (tough not all seem to understand it, like KM) with stored procedures - they sometimes turn into REALLY bad query plans. – Seavey

There's another option on 2008: 'option (optimize for unknown)'. It avoids the 'first use' issue and does not involve a recompile every time. See blogs.msdn.microsoft.com/sqlprogrammability/2008/11/26/… – Familiarize

Answer 2

parameter sniffing google it. try this, which will "remap" the input parameters to local variables to prevent SQL Server from trying to guess the query plan based on parameters:

ALTER PROCEDURE [dbo].[spGetPOIs]
    @lat1 float,
    @lon1 float,
    @lat2 float,
    @lon2 float,
    @minLOD tinyint, 
    @maxLOD tinyint,
    @exact bit
AS
BEGIN
DECLARE @X_lat1 float,
    @X_lon1 float,
    @X_lat2 float,
    @X_lon2 float,
    @X_minLOD tinyint, 
    @X_maxLOD tinyint,
    @X_exact bit



    -- Create the query rectangle as a polygon
    DECLARE @bounds geography;
    SET @bounds = dbo.fnGetRectangleGeographyFromLatLons(@X_lat1, @X_lon1, @lX_at2, @X_lon2);

    -- Perform the selection
    if (@exact = 0)
    BEGIN
        SELECT [ID], [Name], [Type], [Data], [MinLOD], [MaxLOD], [Location].[Lat] AS [Latitude], [Location].[Long] AS [Longitude], [SourceID]
        FROM [POIs]
        WHERE
            NOT ((@X_maxLOD  [MaxLOD])) AND
            (@bounds.Filter([Location]) = 1)
    END
    ELSE
    BEGIN
        SELECT [ID], [Name], [Type], [Data], [MinLOD], [MaxLOD], [Location].[Lat] AS [Latitude], [Location].[Long] AS [Longitude], [SourceID]
        FROM [POIs]
        WHERE
            NOT ((@X_maxLOD  [MaxLOD])) AND
            (@bounds.STIntersects([Location]) = 1)
    END

END

Campbellbannerman

I tried that already, and it didn't seem to make a difference. I'll have to give it another go. – Lactation

Answer 3

I had a similar problem and it was related with indexes.
Rebuilding them help the SP to run fast again.

I found the solution here

USE master;
GO

CREATE PROC DatabaseReIndex(@Database VARCHAR(100)) AS 
BEGIN
  DECLARE @DbID SMALLINT=DB_ID(@Database)--Get Database ID
  IF EXISTS(SELECT * FROM tempdb.sys.objects WHERE name='Indexes') 
  BEGIN --Delete Temp Table if exists, then create
    DROP TABLE TempDb.dbo.Indexes
  END

CREATE TABLE TempDb.dbo.Indexes(IndexTempID INT IDENTITY(1,1),SchemaName NVARCHAR(128),TableName NVARCHAR(128),IndexName NVARCHAR(128),IndexFrag FLOAT)
EXEC ('USE '+@Database+';
INSERT INTO TempDb.dbo.Indexes(TableName,SchemaName,IndexName,IndexFrag)
SELECT OBJECT_NAME(ind.OBJECT_ID) AS TableName,sch.name,ind.name IndexName,indexstats.avg_fragmentation_in_percent
FROM sys.dm_db_index_physical_stats('+@DbID+', NULL, NULL, NULL, NULL) indexstats
INNER JOIN sys.indexes ind ON ind.object_id = indexstats.object_id AND ind.index_id = indexstats.index_id
INNER JOIN sys.objects obj on obj.object_id=indexstats.object_id
INNER JOIN sys.schemas as sch ON sch.schema_id = obj.schema_id
WHERE indexstats.avg_fragmentation_in_percent > 10 AND indexstats.index_type_desc<>''HEAP''
ORDER BY indexstats.avg_fragmentation_in_percent DESC')--Get index data and fragmentation, set the percentage as high or low as you need

DECLARE @IndexTempID BIGINT=0,@SchemaName NVARCHAR(128),@TableName NVARCHAR(128),@IndexName NVARCHAR(128),@IndexFrag FLOAT
SELECT * FROM TempDb.dbo.Indexes --View your results, comment out if not needed...

-- Loop through the indexes
WHILE @IndexTempID IS NOT NULL 
  BEGIN
    SELECT @SchemaName=SchemaName,@TableName=TableName,@IndexName=IndexName,@IndexFrag=IndexFrag FROM TempDb.dbo.Indexes WHERE IndexTempID=@IndexTempID
    IF @IndexName IS NOT NULL AND @SchemaName IS NOT NULL AND @TableName IS NOT NULL 
    BEGIN
      IF @IndexFrag<30. 
      BEGIN --Low fragmentation can use re-organise, set at 30 as per most articles
        PRINT 'USE '+@Database+'; ALTER INDEX ' + @IndexName + N' ON ' + @SchemaName + N'.' + @TableName + N' REORGANIZE'
        EXEC('USE '+@Database+'; ALTER INDEX ' + @IndexName + N' ON ' + @SchemaName + N'.' + @TableName + N' REORGANIZE')
      END
    ELSE 
      BEGIN --High fragmentation needs re-build
        PRINT 'USE '+@Database+'; ALTER INDEX ' + @IndexName + N' ON ' + @SchemaName + N'.' + @TableName + N' REBUILD'
        EXEC('USE '+@Database+'; ALTER INDEX ' + @IndexName + N' ON ' + @SchemaName + N'.' + @TableName + N' REBUILD')
      END
    END

    SET @IndexTempID=(SELECT MIN(IndexTempID) FROM TempDb.dbo.Indexes WHERE IndexTempID>@IndexTempID)
  END
END

DROP TABLE TempDb.dbo.Indexes

GO

Neill