SQL Server Wait_Type full description table

In preparing for another post to continue identifying blocking within your database, I found I did not have a full description for each wait_type, so I took the data from MSDN and created a table that holds all the full descriptions of each wait type, original data here. This was taken from SQL Server 2012 material, your mileage may vary in other versions.

Here is the table I used, very simple:

CREATE TABLE dbo.Wait_Type_Description
   (
   Wait_type VARCHAR(50) NOT NULL,
   Wait_Type_desc VARCHAR(8000) NOT NULL
   )  

and I am joining it like this, again super simple example:

SELECT *
   FROM sys.dm_exec_requests der
       JOIN dbo.wait_Type_description wtd
       ON der.wait_type = wtd.wait_type

All 345 INSERT statements can be found here, sorry this is obnoxious to look at you may download the script or cut and paste it below.

Are your customers abusing you?

Do you have Customers that you dread talking to? You know the types, they are not happy, ever, and have a seemingly endless stream of new requirements (read as: demands) of you and your team. While this post is geared toward those of us who have or are currently working in a large enterprise where database teams and IT specifically work 'for' other internal teams, it applies to everyone with customers or users.
In a large company, you often don't get to 'charge' for your services, the end users get to create requirements and hand them to you, and you don't get to make any requests of them in response, in essence your services are "free". Whenever you offer something for free, you are going to get a somewhat irrational response, people are really effected by the word or concept of "free". Dan Ariely writes about this concept in his book "Predictably Irrational", he shows us through experimentation that when the price is free, the demand will increase incredibly fast, and at the expense of reasonable decision making. I think this holds true in most transactions, whether its free cell phones or free software, if it's free demand will increase beyond what could be expected from otherwise rational consumers. to quote Dan Ariely (@danariely):

... First of all, let me say that there are many times when getting FREE! items can make perfect sense. If you find a bin of FREE! athletic socks at a department store, for instance, there is no downside to grabbing all the socks you can. The critical issue arises when FREE! becomes a struggle between a free item and another item - a struggle in which the presence of FREE! leads us to make a bad decision. *

I have never seen this concept more ably demonstrated than I have working in a large corporate IT department. There are other examples, for example this post on SQLServerCentral's forums is a great example of exactly the same thing happening but in a smaller company

Amazing what a little regulation does for requirements! After looking into PCI requirements, I also found that CSC (or other authentication methods) can never be stored. So when I brought this to the attention of the department head and suggested we look into our options with the lawyer, he quickly rescinded that particular requirement. Turns out, the CSC is not required by our CC processing software, that requirement was just put there "just in case we ever needed it."  (full thread here)

 This is a great example, since there was no cost involved, an unnecessary and in this case legally precarious, requirement was thrown in 'just in case'. For those suffering under the pile of "nice to haves" and "just in case" requirements and constant revisions, what can be done? I have found that even implementing a nominal cost for IT services can have a big impact on the way your customers will view your services. When it comes to nominal cost, it does not have to be monetary, it can be additional regulatory overheard, legal or financial risk, or an unacceptable increase in time lines that brings the needed 'costs' into the equation. The example from the recent forum thread above did not involve any financial costs, the cost was in the potential legal exposure, and the cost was too great. So how does one go about implementing a cost structure in a corporate environment that realistically does not allow you to actually charge money for your interdepartmental services?
If there is a cost center or ledger you can charge, even a nominal amount, then do it. Having to reconcile charges made between departments really isn't money being 'spent' by the company, however having to justify the expense is usually costly enough in terms of hours and headaches to really give someone pause as to whether the expenditure is necessary. I have seen this implemented and work with surprising effectiveness, if the only rationale for requesting something is "why not", then all you need to do is supply that reason. If the concept of charging for a service won't work in your workplace, then charge your customers in a different way, with time. If a requirement is changed or added last minute, address it with the requester by showing him/her what else you are working on and what will suffer delays because of this change, your charging them the opportunity cost of this new item. By having to look over everything they are asking of you and weigh the new requirement against the others, value judgments will be made and unimportant items eliminated quickly.

If you are dealing with a customer or client that is not internal, but are actual customers in the true sense of the word, then it can be trickier to handle. as evidenced here on TechCrunch in a post by @JasonKincaid
there can be a large cost to FREE! promotions, including the overhead to support additional users both with bandwidth/hardware/storage type needs, as well as the user support requirements of exponentially growing a user base that needs support but has paid you nothing in return .In this case there may be a simple answer, stop giving your app away free and you wont have these problems, but if these freeloaders are also paying customers of another of your products or services, alienating them could be costly. You will need or determine whether the traffic and 'exposure' your brand or product is getting has any payoff down the road.

So when you are considering your next set of requirements or thinking of offering a service or product for free, ask your potential customers or users, what is this worth to you? If they can't say, be prepared to propose a value yourself and be confident in asking for compensation for your hard work, and both you and your consumer will be satisfied that value is being exchanged for value.

*Ariely, Dan Predictably Irrational. New York: Harper Perennial 2010
Predictably Irrational, Revised and Expanded Edition: The Hidden Forces That Shape Our Decisions

If you found this interesting or useful, please click the +1 to share (it's FREE!) -- thank you

Deadlock and blocking in SQL Server - Part 2

With SQL Server 2005 a set of dynamic management views were introduced to allow you better visibility into what exactly was happening within the instances of SQL Server you manage or use. The below script takes the outdated sys.sysprocesses and fn_get_sql functions (which still work but are deprecated) and replaces them with new DMVs sys.dm_exec_requests and sys.dm_exec_sql_text.
Additionally it more narrowly defines the SQL statement being executed by both the blockee and blocker by trimming down the full text of the script or procedure and only returning the actual code currently executing. This is done by using the statement_start and statement_end offset values in the sys.dm_exec_requests view.
There is a lot more information available on what currently executing statements are doing within your instance, but this script will get you some of the most important information about who is blocking who quickly. I will expand on this again in the near future with more details about what resources are being contested, and other metrics.

/*
   Created: October 2012
   Author: SQLWorks blog
   URL: http://www.SQLWorks.blogspot.com
   Changelog:
   10/17/2012 - replaced sysprocesses with new DMV sys.dm_exec_requests
              - replaced fn_get_sql with new DMV dm_exec_sql_text
              - Used SUBSTRING and statement_offet values to determine exact statement that               is currently executing
              - instead of returning entire command or proceure
*/

DECLARE @sql_handle TABLE(blockee_spid INT, blockee_cmd VARCHAR(MAX), blocker_spid INT, blocker_cmd VARCHAR(MAX))
INSERT INTO @sql_handle
   SELECT sp1.session_id AS blockee_spid,

           SUBSTRING(blockee_cmd.TEXT,
               CASE WHEN sp1.statement_start_offset >= 0 THEN sp1.statement_start_offset
                   ELSE 0 END,
               CASE WHEN (sp1.statement_end_offset-sp1.statement_start_offset) >= 0 THEN                      (SELECT LEN(TEXT) FROM sys.dm_exec_sql_text(sp1.sql_handle))
                   ELSE (sp1.statement_end_offset-sp1.statement_start_offset) END),

           sp1.blocking_session_id AS blocker_spid,

           SUBSTRING(blocker_cmd.TEXT,
               CASE WHEN sp2.statement_start_offset >= 0 THEN sp2.statement_start_offset
                   ELSE 0 END,
               CASE WHEN (sp2.statement_end_offset-sp2.statement_start_offset)<= 0 THEN          (SELECT LEN(TEXT) FROM sys.dm_exec_sql_text(sp2.sql_handle))
                   ELSE (sp2.statement_end_offset-sp2.statement_start_offset) END) AS blocker_cmd

       FROM sys.dm_exec_requests sp1
           JOIN sys.dm_exec_requests sp2
           ON sp1.blocking_session_id = sp2.session_id
       CROSS APPLY sys.dm_exec_sql_text((sp1.sql_handle)) AS blockee_cmd
       CROSS APPLY sys.dm_exec_sql_text((sp2.sql_handle)) AS blocker_cmd
       WHERE sp1.blocking_session_id <> 0

SELECT *FROM @sql_handle

Using the same test data from Part 1 of this post, I added a second simple SELECT query that would be blocked by the update loop and then ran the above statement. The results show the way that the statements are blocked by each other.

What you can see here is that spid 52 (the update loop) is blocking spid 56 (the first SELECT) and then spid 56 is blocking spid 53 (the second SELECT). These results are a bit hard to look at, so in the next post I will modify how they are displayed and get a more usable result set returned.

Go back to Part 1 - blocking and deadlocks

If you found this useful please click the +1 to share - its FREE!

Deadlocks and blocking in SQL Server

As a DBA you are probably faced with the issue of users blocking each other's queries in your databases and potentially deadlocks between queries as well. Here is how I identify and resolve those issues. This method is simple and quick and probably all you need for most situations, there are much more elaborate ways of tracing and troubleshooting these, but for user queries and in-the-moment identification of issues, this way almost always works.

First, the difference between blocking or contention and a deadlock in simple terms, blocking is when one spid or query blocks another from using a resource or placing the lock it needs on said resource, and a deadlock is when two users or queries each holds a resource the other needs next and neither will let go, so the engine chooses a 'victim' and ends the standoff by killing one query. Again this is really simplified but its all you need to know to get started. Now how do we identify when each scenario is happening?

Lets set up some test data first:

/* Created: October 2012
   Author: SQLWorks Consulting blog
   URL: http://www.SQLWorks.blogspot.com
*/
CREATE TABLE [dbo].[EmployeePerf](
   [EmployeeID] [int] NOT NULL,
   [EmployeeName] [varchar](25) NOT NULL,
   [Sales] [int] NULL,
   [TimePeriod] [int] NULL
)
-- test data
INSERT INTO dbo.EmployeePerf VALUES(1,'Tony',100,'1')
INSERT INTO dbo.EmployeePerf VALUES(1,'Tony',300,'2')
INSERT INTO dbo.EmployeePerf VALUES(1,'Tony',200,'3')
INSERT INTO dbo.EmployeePerf VALUES(1,'Tony',150,'4')  

This is just a table I had set up already on my server to test something else, so I am reusing it, the contents of the table are not really relevant to the discussion on database contention.

So now we need two queries, one to block the other for our testing:
-- Looping update to create a lock for demonstration
DECLARE @sales INT = 100
BEGIN TRAN
WHILE @sales < 500000
BEGIN
   UPDATE [dbo].[EmployeePerf]
   SET sales = @sales
       WHERE Employeename = 'tony' AND timeperiod = 1

   SET @sales = @sales +1
CONTINUE
END

COMMIT TRAN

and in a second window, you will run this concurrently and it will be blocked by the first:

-- select statement that will be blocked by above update statement for demonstration
-- Run this in a different query window!
SELECT *
   FROM dbo.EmployeePerf

So now that we have blocking occurring, what do we do about it? Well first I always like to know who is blocking who, and the commands being attempted by each party. To find that out there are several methods, here is the one I use in SQL Server 2012:

DECLARE @sql_handle TABLE(blockee_spid INT, blockee_cmd VARCHAR(MAX), blocker_spid INT, blocker_cmd VARCHAR(MAX))INSERT INTO @sql_handle
   SELECT sp1.spid AS blockee_spid, blockee_cmd.TEXT AS blockee_cmd, sp1.blocked AS blocker_spid, blocker_cmd.TEXT AS blocker_cmd
       FROM sys.sysprocesses sp1
           JOIN sys.sysprocesses sp2
           ON sp1.blocked = sp2.spid
       CROSS APPLY sys.dm_exec_sql_text((sp1.sql_handle)) AS blockee_cmd --updated
       CROSS APPLY sys.dm_exec_sql_text((sp2.sql_handle)) AS blocker_cmd --updated
       WHERE sp1.blocked <> 0

SELECT *FROM @sql_handle  

This uses the sys.sysprocesses system view to see what is blocking and being blocked and then calls the fn_get_sql function for each to determine what code is being run. If you run the two queries above, then run this script in a third window you will get the below results:

As you can see, it provides the spid of the blocker and blockee, as well as the commands being run by each. There is so much more to delve into on this subject but this will get you moving, and give you a quick way to answer the recurring user question we all love "Who is blocking my query?!?!"


-- EDIT - updated with suggestions from @AaronBertrand of SQLBlog.com
-- changed fn_get_sql to sys.dm_exec_sql_text

-- EDIT - changed formatting of code so comment lines didn't run over code
  -- EDIT - again fixed weird formatting in the code snippets

coming tomorrow, Part 2 - new version using dm views

If you found this useful, click the +1 or Twitter buttons to share -- thanks

SQL code format for web pages

Just found this piece of genius: http://extras.sqlservercentral.com/prettifier/prettifier.aspx 

Paste your code and it will format it to appear like it does in SSMS, and supply you with the HTML. I know there are many, many sites that offer this, but this one actually gets it right according to SSMS colorization. Thank you SQL Server Central.