NOT IN explained

Whenever the column returned by the subquery contains a NULL value, NOT IN will always return an empty set. Filtering out the NULL values in the subquery will solve this problem.

select
    *
from
    table1
where
    col1 not in(select col2 from table2 where col2 is not null)

Read SQL Server: JOIN vs IN vs EXISTS - the logical difference for a detailed explanation.

Corrupt index(es) in msdb

I was trying to delete a SQL Job using sp_delete_job. It continued to fail with the message:

Unable to find index entry in index ID 1, of table {object_id of table}, in database 'msdb'. The indicated index is corrupt or there is a problem with the current update plan. Run DBCC CHECKDB or DBCC CHECKTABLE. If the problem persists, contact product support.

I found this blog post and the script in it repaired the msdb. 

(The command that fixed it is DBCC CHECKDB('msdb', REPAIR_REBUILD), but there is quite a bit of preparation that needs to be done before this command will execute successfully.  See the blog post I referred to.)

Don't be greedy in T-SQL queries

I am working on improving the performance of some stored procedures.  The queries all make use of index seeks and index scans to get to the data, so, nothing I can improve there.
The procs call other procs which call some more procs.  A join to a large table (+/-2million records) gets repeated in several of these procs.

So, I changed the deepest nested proc to return all the data which will be needed by the other procs.  I removed the repeated join; and the performance problem is solved.

So, two very important tips for performance tuning queries:

1. Only return the records you are interested in; and
2. Only touch a record once if possible.

CATCH block not executing after an error in the TRY block

Today we had trouble with the CATCH block in a T-SQL script which did not execute after an error in the TRY block. It turned out that there are times when your TRY block can fail and the CATCH block will get bypassed altogher.

Click here or on the title of this article to see the article by Alexander Kuznetsov about exactly this problem.

Our problem was a compile error: a column name we referenced in a join statement no longer existed.

Protocol error in TDS stream

Yesterday I struggled with the following issue in SSIS: I specified a stored proc as the SQL command text in an OLE DB Source in a data flow. When I clicked the Preview button the data I expected was returned. The Columns were also picked up successfully in the source editor.

But everytime I ran the package it failed with the following errors:

SSIS Error Code DTS_E_OLEDBERROR. An OLE DB error has occurred. Error code: 0x80004005.
An OLE DB record is available. Source: "Microsoft OLE DB Provider for SQL Server" Hresult: 0x80004005 Description: "Protocol error in TDS stream"...

SSIS Error Code DTS_E_PRIMEOUTPUTFAILED. The PrimeOutput method on component "OLE DB Source" (1135) returned error code 0xC0202009. The component returned a failure code when the pipeline engine called PrimeOutput(). The meaning of the failure code is defined by the component, but the error is fatal and the pipeline stopped executing. There may be error messages posted before this with more information about the failure...

A component has returned from its PrimeOutput call. : 1135 : OLE DB Source
SSIS Error Code DTS_E_THREADFAILED. Thread "SourceThread0" has exited with error code 0xC0047038. There may be error messages posted before this with more information on why the thread has exited...

SSIS Error Code DTS_E_THREADCANCELLED. Thread "WorkThread0" received a shutdown signal and is terminating. The user requested a shutdown, or an error in another thread is causing the pipeline to shutdown. There may be error messages posted before this with more information on why the thread was cancelled...

SSIS Error Code DTS_E_THREADFAILED. Thread "WorkThread0" has exited with error code 0xC0047039. There may be error messages posted before this with more information on why the thread has exited...

After searching the whole day for a solution I finally stumbled accross this post by Adam Machanic. At first sight it did not fit my problem, but out of desperation I decided to give it a bash; and it worked.

I am not going to repeat what he said in his post because he did a brilliant job of it. But in short: tables, views and table-valued UDFs have explicitly defined output columns; and stored procs don't. So SSIS will have guaranteed success in retrieving the correct output columns from tables, views and table-valued UDFs, while a stored proc can return different resultsets based on the input parameters.

As I said before, this does not fit my problem since SSIS did pick up the correct output columns, but I suspect that it missed some crutial metadata which wasn't available in the stored procedure.

Get all the dates in a range

Use this T-SQL script to get all the dates in a specified date range.

Please note that it will only work in SQL Server 2005 or later.

SET FMTONLY

When you want to retrieve the metadata for a table, view, stored procedure, etc without having any records returned, use SET FMTONLY ON. Only the column data will be returned.

Get the SQL Server Instance name from T-SQL

To get the SQL Server instance name with T-SQL use one of the following two statements:

• select @@SERVERNAME
• select SERVERPROPERTY('SERVERNAME')

Important: The SERVERNAME property of the SERVERPROPERTY function uses the computer's current network name. So, if the computer (on which the SQL Server instance is installed) had its network name changed after the SQL Server instance was installed, the SERVERPROPERTY function will pick up the new name - @@SERVERNAME will not! @@SERVERNAME will use the network name of the computer at the time when the instance of SQL Server was installed.

Click here to see more properties of the SERVERPROPERTY function.

Get the names of all tables used in a SQL Server view

I want to be able to quickly identify the tables which make up a SQL Server view. I know I can just open the script and look at it, but I have the following problems:

1. All the scripts are one liners for some strange reason.
2. And some of the views contain large numbers of tables in the joins.

So I came up with this script. It will work in SQL Server 2000 and 2005.

Take note: This script is by no means foolproof, and I haven't tested it very extensively. I am even ignoring the tables' schemas altogether. And if a view and a table would have the same name my script wouldn't know the difference.
But hey, it returns the information I want, so I am happy.

I didn't use the system view INFORMATION_SCHEMA.VIEWS because it truncates a view's T-SQL script in the VIEW_DEFINITION column when the script is too long.

I rather used syscomments because it spreads a view's script over several records when the script is too long for the text column. This enabled me to search the whole script for tables names.
I am not sure if syscomments will split a script in the middle of a word when the script is too long for the text column. If that would be the case I could miss some table names because they can be split.
(syscomments is a system view in the sys schema in SQL Server 2005. In SQL Server 7.0 and 2000 it is a system table in the dbo schema.)

Undocumented Stored Procedures

Apparently there are extended stored procedures which are not documented in the Books Online. In this post I want to create a list of those stored procs with links to articles about how to use them. So far I have only one proc, but I will update this post as soon as I come across more.

• xp_ReadErrorLog (SQL Server 2005)

I don't think this one is an extended proc, but also undocumented:

• EXEC sp_MSForEachTable 'EXEC sp_spaceused [?]' - It allows one to execute a T-SQL statement against each table in a database.

Best Practices for Troubleshooting Slow Running Queries

A colleague forwarded this article to me. Although the article contains very useful and valid tips, he pointed out that it does not discuss poorly written T-SQL as a possible cause for slow running queries.

Poorly written T-SQL is probably such a huge topic that it deserves an article of its own.

Anyhow, I hope you can use some of the troubleshooting tips.

SQL Server 2005 ROW_NUMBER() function

Here's an example of how to use the row_number() function which became available in SQL Server 2005.

Click here for a detailed description of the syntax.
Click here for a Google search on this topic.

Run this script against the AdventureWorks sample database in SQL Server 2005. See how the row_number() function numbers the line items per sales order:

select
SalesOrderNumber,
LineItemNo = row_number() over(partition by SalesOrderNumber order by SalesOrderDetailID),
Product = p.Name,
OrderQty,
UnitPrice,
UnitPriceDiscount,
LineTotal
from
Sales.SalesOrderHeader h
inner join
Sales.SalesOrderDetail d
on
d.SalesOrderID = h.SalesOrderID
inner join
Production.Product p
on
p.ProductID = d.ProductID

Change a T-SQL variable in an 'update' statement

declare @message varchar(1024)
set @message = 'New message'

update MyTable
set Column1 = 'New Value',
Column2 = @message,
@message = 'Message updated successfully'
where [Some condition]

select @message

If one or more records have been updated with the update statement, the @message variable will contain "Message updated successfully".
If no record was updated the variable will contain the original text "New message".