http://www.developer.com/

Back to article

SQL Profiler Tips and Tricks


March 16, 2005

If you're a developer who works with SQL Server, you probably already know about SQL Profiler, the graphical tool that lets you capture and analyzer SQL Server events in real time. If you're not familiar with this tool, check out my previous article, "Introduction to SQL Profiler". This time around, I'm going to drill a bit deeper into this useful tool, offering you ten more bits of SQL Profiler that you might not have noticed already.

1. Using Existing Templates

When you fire up SQL Profiler and tell it to create a new trace, it pre-selects some events, data columns, and filters for you. Before you start fine-tuning this selection, you should know that the SQL Server team has already saved some useful starting points for you in the form of trace templates. Instead of creating a new trace, select File, Open, Trace Template. SQL Server ships with trace templates for various purposes including simply counting stored procedures, tuning your SQL statements, and measuring the duration of T-SQL statements.

2. Creating Your Own Templates

Of course, the built-in trace templates won't be perfect for everything you want to do. Sooner or later you'll find yourself carefully crafting a SQL Profiler trace that has the exact combination of events, filters, and data columns that you need to diagnose some common problem within your own organization. When you do, you can stash this combination away for the future as a trace template of your own. Just stop the trace and select File, Save As, Trace Template and assign it a memorable name. The next time you need the same combination you can open up the saved template and have it instantly available.

3. Saving to a Table

One of the nice things about the SQL Profiler engine is that you can either capture data for interactive analysis right on screen, or save it for later inspection - and you don't have to decide up front which of those things you're going to do. Better yet, you can save a trace in the most natural possible place: right in a SQL Server table! When you've created a trace that warrants keeping around for future analysis, stop the trace and select File, Save As, Trace Table. You'll be presented with the Connect to SQL Server dialog box so that you can choose the server where you want to save the trace (this doesn't have to be the same server that you're profiling). Then select a database and either select an existing table or type the name of a new table, as shown in Figure 1. Click OK to save the trace.

Saving a trace to a SQL Server table

4. Replaying a Trace

Saving your traces to a table enables one of the other exciting features of SQL Profiler: replaying traces. Select File, Open, Trace Table and choose a server to connect to that has a saved trace table. Open the trace table. Now you'll find that the commands on the Replay menu are active. Choose Replay, Start and SQL Profiler will let you choose a server to be the target of the replayed activity. Figure 2 shows the options that you can set for a replay.

Replaying a saved trace

Why would you want to replay a trace? Suppose you're debugging a problem with one of your servers - say, clients are deadlocking when running a particular set of queries. You can run SQL Profiler to capture a trace of the client activity that includes the deadlocks, and then rework the stored procedures on the server that you think are causing the deadlock. Replay the stored trace, and you can see whether your fixes were effective in preventing the problem from happening again.

To replay a trace, SQL Server must have certain event classes and data columns in the trace. The easiest way to make sure you have the minimum set in your trace is to start with the SQLProfilerTSQL_Replay trace template, which is one of the ones that's installed with SQL Server.

5. Using Breakpoints

When you're replaying traces, you can use some standard debugging tools to view selected events in slow motion. Place the cursor on any line in the trace and then click F9 to set a breakpoint (or click F9 a second time to clear an existing breakpoint). Then you can click F5 to run the trace to the next breakpoint. At that point you can use the F10 key to execute the trace one statement at a time. Alternatively, you can use Ctrl+F10 to execute all statements up to the current cursor location.

6. Locating Deadlock Causes

If you're having intermittent deadlock problems, it can be tough to figure out where they're coming from. This is especially true if your server is busy: how do you even spot the deadlocks doing by? SQL Profiler can help you here. Set up a trace and in the Events selection expand the Locks group, then select the Lock:Deadlock and Lock:Deadlock Chain events. If you monitor for these two events, SQL Profiler will produce a trace that contains details on just the deadlocks on your server. Record whatever identifying information you want - for example, the application name, logon name, and so on - and you'll be well on your way to figure out where the culprits are.

7. Auditing Logins

How about tracking the user activity on your SQL Server? You can use SQL Profiler for this too. Again, the key lies in properly choosing the events that you profile. Set up a trace that monitors the Security Audit:Audit Login and Security Audit:Audit Logout events, and send it to a trace table. Then you'll have a persistent record right in your database of who was using the database and when they were using it.

8. Watching for Table Scans

Another good use of SQL Profiler is to find queries that are causing table scans on your server. You want SQL Server to be using indexes to find the data that your users require - not to be looking through every row! To check, turn on a trace with the MISC:Execution Plan event. This will capture the query execution plan for every query. Then look over these plans for any that include a "Table Scan" or "Clustered Index Scan" (which indicates that the server is scanning all rows in the index, not that it's using the index to find a particular row). You can then examine those particular queries in more detail to see whether adding additional indexes to your database's tables could make them more efficient in the future.

9. Using Keyboard Shortcuts

Like any other Windows application, SQL Profiler supports keyboard shortcuts for more efficient use. If you use this tool frequently, you'll probably want to memorize some of the most useful ones:

  • Ctrl+Shift+Delete to clear the current trace window
  • Ctrl+F to open the Find dialog box
  • F3 to find the next match
  • Shift+F3 to find the previous match
  • Ctrl+N to open a new trace
  • F5 to start a replay
  • Shift+F5 to stop a replay
  • F9 to toggle a breakpoint
  • F10 to single-step
  • Ctrl+F10 to run to the cursor

10. Choosing Data Columns

Finally, remember that you don't always have to accept the default data columns for your traces. You might find that SQL Server suggests too much or too little data for your tastes. In many cases, the NT User Name and application name will be irrelevant for troubleshooting SQL issues, and you can remove them from the data to avoid cluttering up the display. On the other hand, if you're tracing events for multiple databases, you'll probably want to throw in the database name. There's a lot of other information available, from the duration of the event to the name of the object affected by the current statement, so take a look at the list before you just blindly click OK.

And Yes, There's More

As you should be able to tell by now, SQL Profiler is a vital tool for diagnosing SQL Server issues of all types. When you're trying to figure out what the heck is going on, especially with a heavily-loaded server, a well-chosen trace can help you pick out just the key events that you need to diagnose a problem. Spend some time to get acquainted with the events that you can trace and the data that you can capture, and you'll find many uses for this tool in the future.


Mike Gunderloy is the author of over 20 books and numerous articles on development topics, and the lead developer for Larkware. Check out his recent book, Coder to Developer from Sybex. When he's not writing code, Mike putters in the garden on his farm in eastern Washington state.

Sitemap | Contact Us

Thanks for your registration, follow us on our social networks to keep up-to-date