Monday, September 26, 2016

A Goal of Speaking at PASS Summit

Everyone has goals or at least they should have goals. As a young professional out of college, I had plenty of goals, but none of them involved SQL Server. I really had no idea of where I wanted my career to go other than working in IT. Finally in 2000, I stumbled upon an opportunity that started me down the career path of a DBA. It didn't take long to realize I enjoy this type of work. Over the next decade, I learned a lot from my coworkers, training classes, and especially my mistakes. Yes, I really did drop the master database by mistake.

But one thing I didn't take advantage of was PASS and its huge community of SQL professionals. I knew about some of events, but it wasn't until 2012 that I got heavily involved. That decision forever changed my career. Since then, I started a blog about SQL Server, attended PASS Summit 2013, attended SQL Cruise 2014, started speaking at SQL user groups and SQL Saturdays, started working at Microsoft, attended SQL Cruise 2015, and most of all met a huge group of people that enjoy working with SQL Server as much as I do.

Eventually, I set a goal to be a speaker at PASS Summit. Well this past week I received notification that I had been chosen as a speaker for PASS Summit 2016. Upon hearing the news, I was extremely excited and humbled, especially considering there are far better speakers with more entertaining topics out there.

There are a lot of people that helped me along the way, but I'd like to really say thank you to my friend Andrew Kelly (b|t). Ever since I met him on board SQL Cruise, he has helped improve in the areas of public speaking, designing presentations, writing abstracts, code-review, and just an all-around great mentor.

However, this good news could not have come a worse time. 2016 has been a rough year for my family's health and it got substantially worse in early September when my wife was diagnosed with cancer. Luckily, doctors caught it very early and her prognosis is good, but we're still not taking any chances.

My good friend Scott, who is a cancer-survivor, told me "cancer is a radical disease and must be treated in radical ways" and to be prepared for any type of news throughout the treatment. I'm confident my wife can beat this and make a quick recovery, but until then I'll have to postpone speaking at PASS Summit. While I enjoy spending time with all of my friends within the SQL community, I love my wife and must focus all of my time to her recovery. There will always be time for another Summit.

Tuesday, August 16, 2016

SQLPSX is Finally Getting Updated

The most current code is now on Github, with the Codeplex version being depreciated. You can read all about the planned updates from Mike Shepard.

The Future of SQLPSX


Tuesday, June 14, 2016

TSQL Tuesday #79 -SQL Server 2016 Launch Discovery Day (aka Hackathon)

The host for T-SQL Tuesday #79 is Michael J. Swart (b|t), and his topic of choice is to write about the new release of SQL Server 2016.

This past weekend I attended the SQL Server 2016 Launch Discovery Day in Raleigh, NC. I have attended several SQL Server launch events over the years, but this one was quite different. While it wasn’t a real launch event, it was marketed as a hands-on event to learn about the newest release of SQL Server. The hands-on part for everyone to breakup into teams of five to solve a pre-determined problem. Basically, this was hackathon for SQL Server, and something I’ve never done before.

We started early in the morning with a few presentations about SQL Server 2016. Kevin Feasel (b|t) spoke about the new features, and SQL Server MVP Rick Heiges (b|t) spoke about a real-world win using columnstore indexes in SQL Server 2016. Just before our lunch break the hack was revealed; the basics of which were simple.

First, you are given a set of data set; just five tables containing the following information.
  • Virtual Chapter membership (with member location)
  • SQL Saturday Registration Data and session/track
  • PASS Membership Data
  • Multi-year Summit Data including title, track, score etc
  • Multi-year Summit Registration data including geographic location of attendee
Then you must design a solution to answer these questions.
  • From how far away do attendees travel to SQL Saturday? Are there any geographic trends to the distances traveled?
  • Does the SQL Saturday session data provide any insight into what sessions are chosen for Summit? Are there any trends in session or topic content that can be established?
  • Are there are geographical insights that can discerned? Do the sessions presented at SQL Saturday help predict popularity or selection of sessions at Summit?
  • Does virtual chapter member data provide any insights into PASS membership growth?
The judges score each solution based on the following criteria.
  • Use of new features in SQL Server 2016
  • Usefulness of the dashboard in Power BI/visualizations for the community
  • Completeness of the solution
  • Innovativeness of solution
  • Bonus points for mobile friendliness
Your solution must be completed in 3 hours.

On paper this all sounds pretty easy, but in practice it was quite hard. I am no BI developer and the other members of my team did not have any expertise in that area either, but we still managed to create a solution and have fun doing so.

The first issue was had was how to combine our development work on the same database. This one was easy…just use Azure. In the span of about 30 minutes, I spun up a new Azure VM with SQL Server 2016 pre-installed, uploaded the database, setup logins, and opened the appropriate ports. I then gave my team members the URL and credentials so they each could connect from their laptops.

One of my team members, Dhruv, wanted to get SQL Server R Services installed to analyze the data set. Machine learning was his specialty, and since R Services is a new feature for SQL Server 2016, we thought this would be a good place to start. However, this proved to be mistake for our team. We spent way too much time trying to get it setup. This was mainly do to the the need to install the R components without an internet connection, or I should say a slow connection. I wish we could have implemented this, because Dhruv had some really good ideas. Even without R Services, he was able to create some nifty reports using Power BI.

One of my other team members, Mike, spent a lot of time trying to understand the data, and how to query it to produce the data for our reports. I’m not sure if this was by design, but let me just say the data set was horrible and every team complained about the time needed to clean it up to the point of being useful. Either way, it was just one of the many problems that we needed to overcome. Most of Mike's code was used in the Power BI dashboard that Dhruv created, but he was also able to write some code that made some good points about membership growth potential; however, we did not have time to build a report for it.

Our team (Team Tiger) finished our solution and presented to the group, but it was clear from the other presentations that we had been over matched. The winning solution was from Team Cheetah who had some unique insights into the data and designed their reports to reflect that detail. Not to mention, their presentation went into a lot of detail about what they had learned.

I really liked the entire event, but I wish that we had more time to work on the solution. Three hours seems like a lot, but after thinking about the challenge for a few days, there were so many more ideas that I came up with. For starters, one of the areas to score points was the use of new SQL Server 2016 features. That basically translates into ANY new feature whether it helps you produce a better looking report or not. With that in mind, I could have done the following.
  • Enable the Query Store to capture performance metrics from our solution.
  • Enabled Row-Level Security so we could the reports show different data based on which user is running it.
  • Spin up additional VMs in different Azure datacenters so I could create a load-balanced Availability Group that would provide high availability and better response time for users closer to each datacenter.
  • Setup Stretch Database for a single table to store older data in Azure.
While none of these things would have improved our presentation using Power BI, they are tasks that could have been easily implemented by me given my skillset. And by implementing them it would have definitely scored us a lot more points for the use of new SQL Server 2016 features. This is the big lesion that I learned from the event…always play to your strengths. Don’t try to learn a new skill in a few hours, just use the ones you already have. It will be a much better use of your time, and will most likely produce a better end result.

As I said, this was my first hackathon of any kind, but now I can’t wait to attend another one; especially one that deals with SQL Server.

Tuesday, May 24, 2016

Database Restore Fails with Msg 3154

Have you ever tried to restore over an existing database only to receive the following error message?

Msg 3154, Level 16, State 4, Line 1
The backup set holds a backup of a database other than the existing 'AdventureWorks2012' database.
Msg 3013, Level 16, State 1, Line 1
RESTORE DATABASE is terminating abnormally.

We could easily avoid this error by using the WITH REPLACE option in the RESTORE DATABASE command, but we want to know why this error occurs and how SQL Server knows they are not from the same family. The answer is the database_guid.

The database_guid is a GUID that is unique to every database; even across servers. If you restore a backup to another server, the restored database will get a new database_guid. You can view this value for every database by looking in the catalog view sys.database_recovery_status.

SELECT DB_NAME(database_id), database_guid
FROM master.sys.database_recovery_status;

If the database_guid is different for every database regardless of which SQL Server it’s on, then how does the restore still know the backup and database are from two different families? The answer is the family_guid.

The family_guid is the original database_guid when the database was first created, and does not change for the life of that database. When you issue a RESTORE DATABASE command, SQL Server will compare the family_guid to the database_guid if you are attempting to restore over an existing database. When those numbers don’t match, you will receive that error. You can get the family_guid from the same catalog view.

SELECT DB_NAME(database_id), database_guid, family_guid
FROM master.sys.database_recovery_status

You can use the following code to pull the family_guid from the backup and then compare it to the database you are trying to overwrite.

-- Build a temp table to store the backup file header information.
IF OBJECT_ID('tempdb..#BackupHeader') IS NOT NULL
    DROP TABLE #BackupHeader;

CREATE TABLE #BackupHeader (
     BackupName nvarchar(128)
    ,BackupDescription nvarchar(255)
    ,BackupType smallint
    ,ExpirationDate datetime
    ,Compressed bit
    ,Position smallint
    ,DeviceType tinyint
    ,UserName nvarchar(128)
    ,ServerName nvarchar(128)
    ,DatabaseName nvarchar(128)
    ,DatabaseVersion int
    ,DatabaseCreationDate datetime
    ,BackupSize numeric(20,0)
    ,FirstLSN numeric(25,0)
    ,LastLSN numeric(25,0)
    ,CheckpointLSN numeric(25,0)
    ,DatabaseBackupLSN numeric(25,0)
    ,BackupStartDate datetime
    ,BackupFinishDate datetime
    ,SortOrder smallint
    ,CodePage smallint
    ,UnicodeLocaleId int
    ,UnicodeComparisonStyle int
    ,CompatibilityLevel tinyint
    ,SoftwareVendorId int
    ,SoftwareVersionMajor int
    ,SoftwareVersionMinor int
    ,SoftwareVersionBuild int
    ,MachineName nvarchar(128)
    ,Flags int
    ,BindingID uniqueidentifier
    ,RecoveryForkID uniqueidentifier
    ,Collation nvarchar(128)
    ,FamilyGUID uniqueidentifier
    ,HasBulkLoggedData bit
    ,IsSnapshot bit
    ,IsReadOnly bit
    ,IsSingleUser bit
    ,HasBackupChecksums bit
    ,IsDamaged bit
    ,BeginsLogChain bit
    ,HasIncompleteMetaData bit
    ,IsForceOffline bit
    ,IsCopyOnly bit
    ,FirstRecoveryForkID uniqueidentifier
    ,ForkPointLSN numeric(25,0) NULL
    ,RecoveryModel nvarchar(60)
    ,DifferentialBaseLSN numeric(25,0) NULL
    ,DifferentialBaseGUID uniqueidentifier
    ,BackupTypeDescription nvarchar(60)
    ,BackupSetGUID uniqueidentifier NULL
    ,CompressedBackupSize bigint
    ,containment tinyint NOT NULL

--SQL Server 2014/2016 have three extra columns in the file header.
ALTER TABLE #BackupHeader
ADD  KeyAlgorithm nvarchar(32)
    ,EncryptorThumbprint varbinary(20)
    ,EncryptorType nvarchar(32);

-- Insert file header info into the temp table.
INSERT #BackupHeader
EXEC ('RESTORE HEADERONLY FROM DISK = ''C:\SQL_Backup\AdventureWorks2012.bak''');

-- Compare the family_guid values.
SELECT DatabaseName,FamilyGUID FROM #BackupHeader;

SELECT DB_NAME(database_id),family_guid FROM sys.database_recovery_status
WHERE database_id = DB_ID('AdventureWorks2012');

As you can see they do not match, which is why we get the error.

There is one case where the database_guid can be the same even for different databases. If you detach a database, make a copy of the MDF and LDF files, and then re-attach those files as a different database name, the database_guid values will be the same.

SELECT DB_NAME(database_id), database_guid, family_guid
FROM sys.database_recovery_status;

As you can see, the database_guid and family_guid are the same for all three copies of the AdventureWorks database.