Blog

 

T-SQL Tuesday is a monthly blog party for the SQL Server community. It is the brainchild of Adam Machanic (b|t) and this month’s edition is hosted by Bert Wagner (b|t) who has asked us to “write about code you’ve written that you would hate to live without.”

Interesting.  I go into my toolbox of code/scripts every day.  All day long.  Oddly, most of the things I’ve written are simple things I don’t want to re-type, or are client specific (some going back to 2006 for a custom log shipping troubleshooting process…).  A great many of them are written by others…most often in my case I use things from Adam, Brent and Pinal.  Some are customized, most not.

Recently, I’ve been getting a lot of performance tuning work, much of which is basically “things are slow…can you fix them?” type of requests.  Most experienced DBAs know that there a few gazillion factors that can lead to this request, and I won’t re-hash them here.

Lets assume, that we’ve optimzed or eliminated the server, disks, network, etc. and were now looking at SQL code and everyone’s favorite – indexes.

I have two scripts I use that give me a quick overview of the type of work SQL Server is doing.   These complement each other, and are used AS A STARTING POINT to locate the low-hanging fruit that can be causing excessive work for the disks and memory of the server.

First off – lets look at any HEAPS and Clustered indexes are getting full scans (this is not perfect info…just a starting point remember):

When I run this in its entirety, I get a section of results for every database, for any scans returned by sys.dm_db_index_usage_stats. I can quickly scroll up and down that result set and eyeball any massive numbers for the database(s) in question, or all of them for the server as a whole.  In a sample server I have I found 2 databases with crazy numbers, so I copied the results for those two into Excel and sorted by the Impact column (which is just user_scans * rows):

As you can see on DatabaseB there is a 27 million row table that gets scanned…a lot.  I know some of this is a daily copy from one DB to another…but not that many, so I need to go find out what else is causing these.  While I’ve been writing this, it has been scanned 2000 more times.  On DatabaseA, I have 3 tables I want to look into…2 of them due to sheer size of the tables and one because it is hit so frequently.  In the SQL Query results, the last_user_scan column shows current date and time for all of these…copying to Excel didn’t give me the time…just the date, but that is just a formatting issue.

Note that you can change the results in the WHERE clause to suit your needs.

 

My next step in this process is to run a Query I based off the Pinal Dave’s (b|t) Missing Index query, which simply gets the Top 5 Missing Indxes for each database, ordered by Impact:

Again, use the WHERE clause(s) to filter as you see fit.  In my case I see only one of the tables  from the FULL SCAN query here, so I already know I may have a good shot at seeing at least a valid place to start looking:

I use the Missing Index script all the time, in a variety of ways.   Also, Unused indexes to see if newly created ones are actually in play and being used by the Query Optimizer.

And that’s all I have to say about that…except: USE THESE AT YOUR OWN RISK AND LEARN WHAT THEY DO FIRST! 🙂

Thanks for reading!

Kevin

Dallas DBAs is a small database administration consulting firm. We don’t have a sales staff. Its just me, a Junior DBA, and my wife to inject reality into my delusions of grandeur from time to time.

We live and die from referral business. Nothing beats a happy client sending you something to put on the site, or post on LinkedIn. Even better when they tell someone else.

So how did I make money off Brent Ozar…one of the best known SQL Server DBA/performance dudes in the world?  Well…without him knowing it of course!

Some months ago I got an email from my friend Russ.  A recruiter had approached him about a 4-6 week full time contract cleaning up some SQL Servers.  Russ is more on the Business Intelligence side of the SQL Server world.  I’m a pure DBA so he forwarded the email to me.  I didn’t have the bandwidth for a 40-hour week, but most DBA gigs I see are not actually working a full 8 hours a day, so I fired off an email to the recruiter offering remote, part-time as a choice.  I figured that would be the end of it.  I told him “Cleaning up SQL Server messes is my specialty, and I can probably do it 3x faster than most DBAs.  Plus I won’t milk the client for extra hours.”

Within an hour, the recruiter was setting up a phone call with the client. That day. They lead with “We’ve read your blog, so we aren’t going to ask you technical questions.” (that was very cool…)

Essentially they told me they had signed up for SQL ConstantCare from Brent Ozar Unlimited, which in a nutshell reviews your servers, looks for bad things and tells you what to fix.  I only knew of this product’s existence prior to the call, but no details. I told them I was well aware of the type of tools that Brent and team had created and had been to a number of talks and classes from Brent so I had a pretty good idea what they were looking at. I also told them that if they were basically asking for someone to come in and apply best practices that made sense for them, then I was the guy.

Done deal, send me a contract Mr. Recruiter Guy.

All of this happened in one business day.  From Brent to Client to Recruiter to Russ to me to Recruiter to Client to Me.

Total billing so far for what the client thought was a 4-6 week effort?  Right at 40 hours, done in my spare morning hours I keep open for simple tasks on smaller clients.

If you have a bunch of servers, no SQL DBA and you need to know where you stand…ConstantCare is a great idea.  Then ping me and I’ll fix them for you.

Thanks for reading!

Kevin3NF

Follow @Dallas_DBAs

Full scans of a HEAP or Clustered index in SQL Server aren’t necessarily a bad thing.  But sometimes they are scanned because the Query Optimizer (“Optimus” here at Dallas DBAs HQ) cannot find a suitable index to use.  Sometimes its just because the table is tiny.  Either way, its worth investigating WHY all the pages are being read…especially if you are having performance issues or memory pressure.

Just for fun, I took the query found in this DBA StackExchange Answer and modified it to fit a client’s need.   Decided to share:

-- base query provided by Jonathan Fite:
-- https://dba.stackexchange.com/questions/210341/number-of-full-table-scans
-- Added sys.Partitions for rowcount, Imapct computed column, Where conditions and Sorting
-- Kevin3NF, https://dallasdbas.com

SELECT S.name AS SchemaName
  , O.name AS ObjectName
  , I.name AS IndexName
  , I.type_desc AS IndexType_Desc
  , IUS.last_user_scan
  , IUS.user_scans
  , P.rows
  , IUS.user_scans * P.rows as [Impact]
FROM sys.objects AS O 
  INNER JOIN sys.schemas AS S ON S.schema_id = O.schema_id
  INNER JOIN sys.indexes AS I ON I.object_id = O.object_id
  INNER JOIN sys.dm_db_index_usage_stats AS IUS ON IUS.index_id = I.index_id AND IUS.object_id = I.object_id
  INNER JOIN sys.Partitions P on p.object_id = I.object_id and p.index_id = i.index_id
WHERE 1=1 
  and I.type IN (0,1) --0=HEAP, 1=CLUSTERED (table)
  and IUS.user_scans > 5
  and p.rows > 1000

Order by 
--IUS.user_scans desc
--p.Rows Desc
IUS.user_scans * P.rows DESC

The top rows are very likely good candidates for an index review.

Edit as you see fit for your environment, especially the WHERE clause. Please leave the headers.

Thanks for reading.

Kevin3NF