Yes, that works too! I’ll edit my post with your suggestion too. Thank you.

Michelle

I actually mispoke (mistyped?) a little. I thought the error message was returned that way because it was called via stored proc, but really you would get the same error message if you just try to force the hint in an ad-hoc query, too.

You receive the error message because Query Optimizer, sometimes correctly and sometimes incorrectly, perceives that it will be unable to fulfill the request using the index you specify. The only case I’ve ever seen this in (so far) has been a filtered index. Hopefully that helps, and good luck!

Michelle

I think that if you explicitly tell the optimizer that your parameter fits into the filter. Instead of:

SELECT DISTINCT myData
FROM dbo.filteredIndexTest
WHERE myDate >= @myDate1;

let us try this:

SELECT DISTINCT myData
FROM dbo.filteredIndexTest
WHERE myDate >= @myDate1
AND myDate>’theThresholdInMyFilteredIndex’

What do you think, Michelle?

However, I am much more interested in what causes this error in the first place. You state, “…since this was a stored procedure, I was receiving the following error message…” implying this is due to sprocs. However, we’ve been using index hints in sprocs for years, and have never received this error. I am curious why I am getting it now all of a sudden on one specific sproc. Do you or any of your readers have insight on that?

Thanks!

IF EXISTS(SELECT * FROM dbo.sysobjects WHERE id = OBJECT_ID(‘dbo.DBMaint_Task’) AND xtype = ‘U’)
BEGIN
DROP TABLE dbo.DBMaint_Task
END
GO

CREATE TABLE dbo.DBMaint_Task(
MaintTaskId int NOT NULL IDENTITY(1, 1),
Interval int NOT NULL,
Command nvarchar(4000) NOT NULL,
ItemWork numeric(10,2) NOT NULL,
CompletedDatetime datetime NULL
CONSTRAINT PK_DBMaint_Task PRIMARY KEY CLUSTERED(MaintTaskID)
)
GO

IF EXISTS(SELECT * FROM dbo.sysobjects WHERE id = OBJECT_ID(‘dbo.p_maint_ScheduleMaintTasks’) AND xtype = ‘P’)
BEGIN
DROP PROCEDURE dbo.p_maint_ScheduleMaintTasks
END
GO

CREATE PROCEDURE dbo.p_maint_ScheduleMaintTasks
@PeriodLength int, – Number of intervals in a period
@IndexRebuildThreshold int = 10 – Maximum LogicalFragmentation value allowed before an index is rebuilt
AS

DECLARE @WorkPerIntervalGoal numeric(10,2)
DECLARE @CurrentInterval int
DECLARE @RowsUpdated int
DECLARE @MainTaskId int

SET NOCOUNT ON

IF @PeriodLength = @IndexRebuildThreshold
AND
ObjectId OBJECT_ID(‘dbo.DBMaint_Task’)

– If we rebuild a clustered index, we don’t need to rebuild
– any nonclustered indexes on the same table
DELETE FROM
rc1
FROM
#ReindexCandidate rc1
INNER JOIN
#ReindexCandidate rc2
ON(
rc1.ObjectId = rc2.ObjectId
)
WHERE
rc2.IndexID = 1 — IndexID = 1 means the clustered index
AND
rc1.IndexID 1

TRUNCATE TABLE dbo.DBMaint_Task

INSERT INTO dbo.DBMaint_Task(
Interval,
Command,
ItemWork)
SELECT
-1 AS Interval,
N’DBCC DBREINDEX(”’ + USER_NAME(OBJECTPROPERTY(rc1.ObjectId, ‘OwnerId’)) + N’.’ + rc1.ObjectName + N”” +
CASE IndexId WHEN 1 THEN N” ELSE N’, ”’ + IndexName + N”” END + N’)’ AS Command,
(SELECT SUM(dpages) FROM dbo.sysindexes si1 WHERE id = rc1.ObjectId) AS ItemWork
FROM
#ReindexCandidate rc1

– Find all tables that we’re not doing a clustered index rebuild on
INSERT INTO dbo.DBMaint_Task(
Interval,
Command,
ItemWork)
SELECT
-1 AS Interval,
N’UPDATE STATISTICS [' + USER_NAME(uid) + N'].[' + so1.name + N'] WITH FULLSCAN’ AS Command,
(SELECT SUM(dpages) FROM dbo.sysindexes si1 WHERE id = so1.id) * 0.40 AS ItemWork
FROM
dbo.sysobjects so1
LEFT OUTER JOIN
#ReindexCandidate rc1
ON
(
rc1.ObjectId = so1.id
AND
rc1.IndexId = 1
)
WHERE
so1.xtype = ‘U’
AND
rc1.ObjectId IS NULL
AND
(SELECT SUM(dpages) FROM dbo.sysindexes si1 WHERE id = so1.id) 0
AND
so1.id OBJECT_ID(‘dbo.DBMaint_Task’)

DELETE FROM
dbo.DBMaint_Task
WHERE
ItemWork = 0

– Break all ties by ItemWork
– This algorithm is arbitrary, but the amount of estimated work it adds is trival
SET @RowsUpdated = 1
WHILE @RowsUpdated 0
BEGIN
UPDATE
mt1
SET
ItemWork = ItemWork + 0.01
FROM
dbo.DBMaint_Task mt1
INNER JOIN
(
SELECT
MAX(MaintTaskId) AS MaintTaskId
FROM
dbo.DBMaint_Task
GROUP BY
ItemWork
HAVING
COUNT(*) > 1
) mt2
ON
(
mt1.MaintTaskId = mt2.MaintTaskId
)

SET @RowsUpdated = @@ROWCOUNT
END

SET @CurrentInterval = 1
WHILE @CurrentInterval = mt1.ItemWork
AND
mt2.Interval = -1
) <= @WorkPerIntervalGoal

SET @RowsUpdated = @@ROWCOUNT

IF @RowsUpdated = 0
BEGIN
UPDATE
mt1
SET
Interval = @CurrentInterval
FROM
dbo.DBMaint_Task mt1
WHERE
mt1.ItemWork = (
SELECT
MAX(mt2.ItemWork)
FROM
dbo.DBMaint_Task mt2
WHERE
mt2.Interval = -1
)
END

SET @CurrentInterval = @CurrentInterval + 1
END

SET NOCOUNT OFF

GO

IF EXISTS(SELECT * FROM dbo.sysobjects WHERE id = OBJECT_ID('dbo.p_maint_ExecuteMaintTasks') AND xtype = 'P')
BEGIN
DROP PROCEDURE dbo.p_maint_ExecuteMaintTasks
END
GO

CREATE PROCEDURE dbo.p_maint_ExecuteMaintTasks
@PeriodLength int = 7, – Number of intervals in a period
@IndexRebuildThreshold int = 10, – Maximum LogicalFragmentation value allowed before an index is rebuilt
@ForceInterval int = NULL, – Manually set the current interval rather than calculating it from the date
@ForceRebuildTaskList int = 0, – If non-zero, force a rebuild of the task list
@RebuildTaskListInterval int = 1 – The interval in which to automatically rebuild the task list
AS

DECLARE @CurrentInterval int
DECLARE @TaskCount int
DECLARE @Command nvarchar(4000)
DECLARE @MaintTaskId int

SET NOCOUNT ON

IF @PeriodLength < 1
BEGIN
RAISERROR(N'@PeriodLength must be greater than zero.', 16, 1)
RETURN
END

IF @ForceInterval IS NOT NULL
BEGIN
IF @ForceInterval @PeriodLength
BEGIN
RAISERROR(N’If not null, @ForceInterval must be less than or equal to @PeriodLength.’, 16, 1)
RETURN
END
END

IF @RebuildTaskListInterval > @PeriodLength
BEGIN
RAISERROR(N’@RebuildTaskListInterval must be less than or equal to @PeriodLength.’, 16, 1)
RETURN
END

IF NOT @IndexRebuildThreshold BETWEEN 0 AND 100
BEGIN
RAISERROR(N’@IndexRebuildThreshold must be between 0 and 100.’, 16, 1)
RETURN
END

SELECT
@TaskCount = COUNT(*)
FROM
dbo.DBMaint_Task
WHERE
CompletedDatetime IS NULL

IF @ForceInterval IS NULL
BEGIN
SELECT
@CurrentInterval = ((YEAR(GETDATE() – 59) * 365 + DATEPART(dy, GETDATE() – 59)) % @PeriodLength) + 1
END
ELSE
BEGIN
SELECT
@CurrentInterval = @ForceInterval
END

IF @TaskCount = 0 OR @ForceRebuildTaskList 0 OR @CurrentInterval = @RebuildTaskListInterval
BEGIN
EXEC dbo.p_maint_ScheduleMaintTasks
@PeriodLength = @PeriodLength,
@IndexRebuildThreshold = @IndexRebuildThreshold
END

DECLARE csr_Task CURSOR FAST_FORWARD FOR
SELECT
MaintTaskId,
Command
FROM
dbo.DBMaint_Task
WHERE
Interval = @CurrentInterval
AND
CompletedDatetime IS NULL

OPEN csr_Task

FETCH NEXT FROM csr_Task INTO @MaintTaskId, @Command
WHILE @@FETCH_STATUS = 0
BEGIN
PRINT CONVERT(nvarchar(30), GETDATE(), 121) + N’: Executing “‘ + @Command + N’”‘
EXEC(@Command)
PRINT ”

UPDATE
dbo.DBMaint_Task
SET
CompletedDatetime = GETDATE()
WHERE
MaintTaskId = @MaintTaskId

FETCH NEXT FROM csr_Task INTO @MaintTaskId, @Command
END

CLOSE csr_Task
DEALLOCATE csr_Task

SET NOCOUNT OFF
GO

Is this latest version of the script that has addressed some of the issues by users (like Derick and SuperCoolMoss pointed about the stats rebuild feature in your defrag script will only complete for one database.)?

Thank you