Monitoring DTS Steps

Part of my work involves a SQL Server 2000 data warehouse,
updated nightly with extracts from an ERP system.  I knew that the DTSs
tended to run long, and I wanted to get a better idea of exactly how much data
we were moving around.

First thing I did was add a table with the following

 [fldEntry] [int] IDENTITY (1, 1) NOT NULL ,
[int] NOT NULL ,
 [fldStepType] [varchar] (5) COLLATE
SQL_Latin1_General_CP1_CI_AS NOT NULL ,
 [fldStepName] [varchar] (50)
[datetime] NOT NULL ,
 [fldNotes] [varchar] (255) COLLATE
SQL_Latin1_General_CP1_CI_AS NULL ,
 [fldPackage] [varchar] (50)
) ON

This is a pretty general structure, and allows me to
add different logging and debugging steps to the DTS.  fldEntry is simply
an incrementing value to help sorting what happened when.  fldStep refers
to the stepf of the DTS package, fldStepType is a short description of the DTS
step.  I use fldStep because someone always asks me what that step is doing
when we’re looking at the numbers, and it’s easier to log it than open the DTS
and find it.  fldStepName is there because I inherited almost all of these
DTSs, and the names don’t always describe what’s going on.  fldTimeStamp is
the execution time of the step, fldNotes are some information I want to include
about the step (usually just a rowcount), and fldPackage is the name of the DTS

One of the early steps for most DTSs is the extract
from the ERP system to an intermediate table.  Depending on the package,
the data in the intermediate table used for comparisons or modified before being
migrated to the production tables.  By adding  a SQL task immediately
after the extract, I could get an exact idea of how many data rows we just

select * from IntermediateTable
into tblDtsLog(fldStep, fldStepType, fldStepName, fldTimeStamp,fldNotes,
fldPackage) values (1,’Begin’, ‘Extract Step 1’,getdate(),@@rowcount, ‘Some

This simly adds one entry to the above table with a
count of how many rows exist in the intermediate table.  
For other steps, I use the Begin and End flags in conjunction with
the getdate() to see how much time elapsed for the processing of a particular
matching or modification step.  By sprinkling these liberally through my
DTSs (some packages have 15-20 of these logging sinks in them), I have excellent
insight into what is running slow, and then we can examine each step and try to
optimize it.

Recently, we used the results of this monitoring to
make a small tweak to one step of one DTS, which resulted in an overall gain of
2 hours of processing time in the overall update!

Copying DTS Packages to a Different Server in SQL Server 2000

Freaky–we were just talking about this yesterday afternoon.

The latest
update came out today, and the lead story is an update to an older article
titled “Copying
DTS Packages to a Different Server
“.  The article has some code samples
to show you how you can copy DTS packages from one SQL Server 2000 to

If you’re lazy, don’t bother with the code.  Instead, there’s a great
GUI tool called DTSBackup
 from  Using
DTSBackup 2000 couldn’t be easier–choose your origin server (supply
credentials if necessary), choose the DTSs you want to copy, choose your
destination server (supply credentials if necessary), add in the owner pwd if
necessary, and click Transfer.  Within minutes, your packages are
transferred.  It’s sweet, and works like a charm.

I use this to copy DTSs from my dev/test server to the prod server, and back
again if necessary.  Layout in the visual designer gets a little goofy, but
that’s the only downside.

Is Dynamic SQL in Your Stored Procedures Vulnerable to SQL Injection?

We all should be familiar with the fact that concatenating user input directly into SQL statements is an open invitation to an SQL Injection attack.  Code such as
MySql = “Select * from Orders where Customer ID='” & txtCustomerId & “‘”
should be avoided.  If you need some more background information on SQL Injection attacks, I am building a reference at
.  This reference will be updated as time goes on–there are a few good references now, and I’ll post update notices to the security section of this blog.

The recommended practice for avoiding SQL Injection attacks is to use parameterized queries or stored procedures (sprocs), where user input is passed as parameters.  Since information in parameters is not treated as executable code, any SQL code conatined in the user input is rendered harmless.  Or is it?  This depends on what you do with that input inside of your sproc.

One of the common functions on a web site is querying a data store.  In advanced searches (those with more than a single input), it would be infeasible to create and mainatin an sproc for every combination of search critera.  Instead, one practice is to create an sproc that dynamically creates the SELECT statement based on the parameters passed to it.  Typically, there is an input parameter for each input on the search form, which is rendered optional by adding “=NULL” after the parameter declaration (e.g., @orderId int=NULL).  Then, the sproc uses a series of statements such as

 select @sql = @sql + ‘ AND order_id=’ + @orderId
to generate the complete SQL statement.  At the end of the sproc, the EXECUTE statement is used to query the database using the dynamically generated SQL statement.

I remember what a revolutionary concept dynamic SQL in an sproc was for me when I was learning to write SQL.  It opened up a whole new way of writing SQL code and handling advanced searches on my websites.  But did you catch the security problem in the previous SQL statement?  I didn’t at first, and in fact, I’ve been making this same security mistake for some time now.  It wasn’t until I finally listened to Kim Tripp on DotNetRocks that I realized the problem (download the show from, and fortunately I only have a few sprocs to rewrite and fix this problem.

Look carefully at the statement again.  It looks like the parameter is being used in the SQL statement, but in reality, the parameter’s value is being concatenated to the SQL statement.  The technique demonstrated above is no better than the technque we dismissed in the first paragraph.

After listening to Kim’s show, I did some digging around, and found an excellent reference on how to handle dynamic SQL in search queries at  In this article, Microsoft Sql Server MVP Erland Sommarskog details ways to use dynamic and static SQL to perform searches that have a number of possible combinations of inputs.

As Erland shows us, the correct way to use dynamic SQL in the situation I presented above is to concatenate another parameter into the SQL statement, as so:

 select @sql = @sql + ‘ AND

We then create a parameter list of these second parameters, as so:

SELECT @paramList = ‘@xorderId’

To finally execute the query, we execute a system sproc named sp_executesql.  As Erland states:

sp_executesql is a system procedure with a very special parameter list. The first parameter is a parameterized SQL statement. The second parameter is a parameter-list declaration, very similar to the parameter list to a stored procedure. And the remaining parameters are simply the parameters in that parameter-list parameter.

Our final statement would end up looking like:

EXECUTE sp_executesql @sql, @paramList, @orderId

And with this technique, our query is safe from malicious user input.  This whole process is outlined in detail in Erland’s article.

Since writing sprocs as outlined in Erland’s article can be tedious, I created a CodeSmith template that will do the work for you.  You only need to input the table you wish to query, and CodeSmith will generate a complete sproc for you.  You can then edit the sproc down, since it will include every column in the table.  You can find the template at

<update 2005-07-06: fixed DNR show link>