SQL Server Stuff

Scriptio available on CodePlex

Bill Graziano — Thu, 05 Jun 2008 14:12:35 GMT

Scriptio is a little utility I wrote when I was unhappy with the scripting features in SQL Server 2005. The biggest problem I faced was that I couldn't script a DROP and CREATE in the same file. I'm pretty sure that's finally fixed. The utility gives you all kind of scripting options now.

The source code was always available on the site but wasn't always easy to find. And there wasn't a good way for people to submit improvements. I finally migrated the code base over to CodePlex. You can download the source code or just the compiled EXE. It should also make it much easier to submit improvements.

I'm not sure what the future of this utility is. I haven't played much with the scripting features in SQL Server 2008. It is a great example of how to use SMO and the SMO scripting component.

A great big Thank You to Riccardo Spagni for prompting me to do this. He also rewrote most of the code so it looks less like something I cooked up in my spare time (which it was) and looks more like an actual application. Thanks!

Scriptio on CodePlex.

Are you attending the SQL Server 2008 Launch in LA?

Bill Graziano — Mon, 25 Feb 2008 13:27:04 GMT

I'll be in Los Angeles for the launch this week. I'll be the PASS representative at the Heroes Community Booth in the Microsoft Pavilion. When you enter the Microsoft Pavilion we'll be all the way back in the left corner. Come by and say Hi if you get a chance. I arrive Tuesday night and leave Thursday morning. Is anyone else reading going to be there?

SQL Server 2008 to RTM in Q3 of 2008

Bill Graziano — Fri, 25 Jan 2008 21:58:58 GMT

Francois Ajenstat posted today that Microsoft expects to release SQL Server 2008 to manufacturing in the third-quarter of 2008.

Microsoft is excited to deliver a feature complete CTP during the Heroes Happen Here launch wave and a release candidate (RC) in Q2 calendar year 2008, with final Release to manufacturing (RTM) of SQL Server 2008 expected in Q3. Our goal is to deliver the highest quality product possible and we simply want to use the time to meet the high bar that you, our customers, expect.

Their previous plan was to RTM in Q2 so it looks like things are slipping a little. Still I'd rather have a little delay than release the product before it's ready.

SQL Server 2005 Best Practices Analyzer Updated (Jan 2008)

Bill Graziano — Sat, 19 Jan 2008 15:40:43 GMT

Microsoft just updated the SQL Server 2005 Best Practices Analyzer. There are over 60 new rules included in this release. You can find more information on the SQLCAT blog.

ReadTrace for SQL Server 2005

Bill Graziano — Thu, 20 Dec 2007 01:59:42 GMT

Microsoft just released an updated Read80Trace that works on SQL Server 2005. The name is simplified to just ReadTrace. ReadTrace is a command-line utility that analyzes a SQL Server trace file and summarizes performance. The new version has a more graphical interface. The original was one of the most helpful performance tuning utilities I've ever used. I expect this one will be just as useful.

I typically use the tool to summarize performance by SQL statement. ReadTrace will "normalize" the SQL statements by replacing constants with placeholders. This lets you identify which SQL statements are using the most resources.

You can read the PSS team's blog entry or just download the utilities. There is some additional functionality for replaying traces.

SQL Server 2008 CTP5 on Friday

Bill Graziano — Thu, 15 Nov 2007 20:41:49 GMT

It looks like SQL Server 2008 CTP5 should be available for download on Friday. It's already in Connect for certain testing groups.

Login Events include Pooled Connections

Bill Graziano — Wed, 31 Oct 2007 13:52:26 GMT

Every now and then I stumble across something and say "I know I knew this before!" In Profiler, the Audit Login event includes regular logins and connection pooling reuse. If you run a trace and see a large number of logins you'll need to include the EventSubClass column to determine if they are pooled connections.

Even more misleading is PerfMon. If you capture the SQLServer:General Statistics - Logins/sec counter you'll see regular logins and connection pooling reuse. I haven't found a way to have PerfMon just show me "real" logins and not connection pooling events. I've started watching User Connections and Logouts/sec to see how many logins are really happening. Logouts/sec doesn't report anything related to connection pooling.

On SQLTeam.com I have roughly twenty logins per second but under one logout per second while the number of connections is stable. That tells me I really have one "real" login per second and nineteen connection pooling reuse events per second.

Read90Trace coming soon!

Bill Graziano — Mon, 24 Sep 2007 03:20:49 GMT

Bob Ward announced in his PASS pre-conference session that Microsoft will be releasing a SQL Server 2005 version of Read80Trace in roughly 30 days. It's currently going through a security review. They're planning to call it ReadTrace instead of Read90Trace though. I don't know what this will do to my plans for more work on ClearTrace. I'll have to wait and see exactly what they release.

Viewing Database Updates through ODBC using Profiler

Bill Graziano — Fri, 24 Aug 2007 15:35:49 GMT

I just finished writing a small data update utility to copy a single row of data from a custom written application into a piece of packaged software. Usually when I do this I turn on Profiler, watch the packaged software insert a row and then copy what they did. This helps me find all the odd little status fields and related tables that aren't obvious from looking at the table. And yes, if the developer provided documentation it would help. They didn't.

A simple login, adding one row and logging out generated over 10,000 SQL:BatchCompleted events since this application uses ODBC cursors for all its data access. This meant I had to wade through 10,000 sp_cursor* statements to find the dozen actual SQL statements that were issued.

Last year I wrote an article on examining SQL Server Trace files. In that article I show examples of what a server-side cursor or SQL statements issued through ODBC look like:

declare @P1 int
set @P1=180150001
declare @P2 int
set @P2=8
declare @P3 int
set @P3=1
declare @P4 int
set @P4=1
exec sp_cursoropen 
  @P1 output, 
  N'select SomeColumn FROM TableName', 
  @P2 output, 
  @P3 output, 
  @P4 output
select @P1, @P2, @P3, @P4
GO

exec sp_cursorfetch 180150001, 2, 1, 1
GO

exec sp_cursorclose 180150001
GO

Today I got the "opportunity" to see how updates worked in this case. Using a similar example an update would look like this:

declare @p1 int
set @p1=1073741854
declare @p2 int
set @p2=180150297
declare @p5 int
set @p5=1
declare @p6 int
set @p6=4
declare @p7 int
set @p7=0
exec sp_cursorprepexec @p1 output,@p2 output,NULL,
	N'SELECT Column1, Column2 FROM TableName WHERE PkColumn = 37',
	@p5 output,@p6 output,@p7 output
select @p1, @p2, @p5, @p6, @p7
GO

exec sp_cursorfetch 180150297,2,1,1
GO

exec sp_cursor 180150297,4,1,N'',@Column1='2007-0660',@Column2='Accident'
GO

exec sp_cursorfetch 180150297,8,-1,1
GO

exec sp_cursorclose 180150297
GO

The sp_cursor statement declares the new values for the columns that are updated. I found some third-party documentation on sp_cursor but didn't find much from Microsoft in very limited searching. Apparently the second parameter indicates whether this is an update or insert. The sp_cursorfetch with the -1 appears to commit the row back to the database.

The sp_cursor statement can be separated from the corresponding statement to create the cursor by hundreds of rows in Profiler. I found myself searching for sp_cursor, finding the cursor number and then searching backwards in the trace to find the declaration of the cursor.

Upgrading to SAS drives

Bill Graziano — Fri, 27 Apr 2007 11:45:13 GMT

We were having some disk drive performance issues at a client recently. They have a 6 drive array running RAID 10. We spent some time running SQLIO against the array. Our standard test was an 8KB random read with a queue depth of eight. The array was giving us 370 I/O operations per second with a latency of 42 ms. We cracked open the case and discovered it was using 7200 RPM SATA drives. We replaced the controller with the same model except that it was designed for SAS (serially attached SCSI) and installed six 15,000 RPM SAS drives. Running the identical test now resulted in 1,980 I/O operations per second with a 7 ms latency. Nice!

Next, we rebuilt the partition using DISKPART.EXE to align the partitions. (Yes, I know the instructions are for Exchange. But that's the best description I've found and that's exactly what we did.) We used a setting of 64 for the alignment. That improved our performance to 2,142 and 6 ms latency. Aligning the partitions increased our I/O throughput by just over 8%.