Convert VS 2010 Database Project to SSDT and Automate Publishing with Jenkins – Part 3/3

Objectives of 3-Part Series:

Part I: Setting up the Example Database and Visual Studio Projects

  • Setup and configure a new instance of SQL Server 2008 R2
  • Setup and configure a copy of Microsoft’s Adventure Works database
  • Create and configure both a Visual Studio 2010 server project and Visual Studio 2010 database project
  • Test the project’s ability to deploy changes to the database

Part II: Converting the Visual Studio 2010 Database and Server Projects to SSDT

  • Convert the Adventure Works Visual Studio 2010 database and server projects to SSDT projects
  • Create a second Solution configuration and SSDT publish profile for an additional database environment
  • Test the converted database project’s ability to publish changes to multiple database environments

Part III: Automate the Building and Publishing of the SSDT Database Project Using Jenkins

  • Automate the build and delivery of a sql change script artifact, for any database environment, to a designated release location using a parameterized build.
  • Automate the build and publishing of the SSDT project’s changes directly to any database environment using a parameterized build.

Part III: Automate the Building and Publishing of the SSDT Database Project Using Jenkins

In this last post we will use Jenkins to publishing of changes from the Adventure Works SSDT database project to the Adventure Works database. Jenkins, formally Hudson, is the industry-standard, java-based open-source continuous integration server.


If you are unfamiliar with Jenkins, I recommend an earlier post, Automated Deployment to GlassFish Using Jenkins and Ant. That post goes into detail on Jenkins and its associated plug-in architecture. Jenkins’ website provides excellent resources for installing and configuring Jenkins on Windows. For this post, I’ll assume that you have Jenkins installed and running as a Windows Service.

The latest available version of Jenkins, at the time of this post is 1.476. To follow along with the post, you will need to install and configure the following (4) plug-ins:

User Authentication

In the first two posts, we connected to the Adventure Works database with the ‘aw_dev’ SQL Server user account, using SQL Authentication. This account was used to perform schema comparisons and publish changes from the Visual Studio project. Although SQL Authentication is an acceptable means of accessing SQL Server, Windows Authentication is more common in corporate and enterprise software environments, especially where Microsoft’s Active Directory is used. Windows Authentication with Active Directory (AD) provides an easier, centralized user account security model. It is considered more secure.

With Windows Authentication, we associate a SQL Server Login with an existing Windows user account. The account may be local to the SQL Server or part of an Active Directory domain. For this post, instead using SQL Authentication, passing the ‘aw_dev’ user’s credentials to SQL Server in database project’s connection strings, we will switch to Windows Authentication. Using Windows Authentication will allow Jenkins to connect directly to SQL Server.

Setting up the Jenkins Windows User Account

Let’s outline the process of creating a Jenkins Windows user account and using Windows Authentication with our Adventure Works project:

  1. Create a new ‘jenkins’ Windows user account.
  2. Change the Jenkins Windows service Log On account to the ‘jenkins’ Windows account.
  3. Create a new ‘jenkins’ SQL Server Login, associated with the ‘jenkins’ Windows user account, using Windows Authentication.
  4. Provide privileges in SQL Server to the ‘jenkins’ user identical to the ‘aw_dev’ user.
  5. Change the connection strings in the publishing profiles to use Windows Authentication.

First, create the ‘jenkins’ Windows user account on the computer where you have SQL Server and Jenkins installed. If they are on separate computers, then you will need to install the account on both computers, or use Active Directory. For this demonstration, I have both SQL Server and Jenkins installed on the same computer. I gave the ‘jenkins’ user administrative-level rights on my machine, by assigning it to the Administrators group.

Create New Jenkins User

Create New Jenkins User

Next, change the ‘Log On’ Windows user account for the Jenkins Windows service to the ‘jenkins’ Windows user account. Restart the Jenkins Windows service to apply the change. If the service fails to restart, it is likely you did not give enough rights to the user. I suggest adding the user to the Administrators group, to check if the problem you have encountering is permissions-related.

Jenkins Windows Service

Jenkins Windows Service

Set Log On Account for Jenkins Windows Service

Set Log On Account for Jenkins Windows Service

Log On Account for Jenkins Windows Service

Log On Account for Jenkins Windows Service

Log On Account for Jenkins Windows Service Granted

Log On Account for Jenkins Windows Service Granted

Setting up the Jenkins SQL Server Login

Next, to use Windows Authentication with SQL Server, create a new ‘jenkins’ Login for the Production instance of SQL Server and it with the ‘jenkins’ Windows user account. Replicate the ‘aw_dev’ SQL user’s various permissions for the ‘jenkins’ user. The ‘jenkins’ account will be performing similar tasks to ‘aw_dev’, but this time initiated by Jenkins, not Visual Studio. Repeat this process for the Development instance of SQL Server.

Jenkins Login Added to Development Instance

Jenkins Login Added to Development Instance

Jenkins User Any Definition on Production Instance

Jenkins User Any Definition on Production Instance

Jenkins User View Definition on Production Instance Database

Jenkins User View Definition on Production Instance Database

SSMS View of Jenkins User

SSMS View of Jenkins User

Windows Authentication with the Publishing Profile

In Visual Studio, switch the connection strings in the Development and Production publishing profiles in both the server project and database projects to Windows Authentication with Integrated Security. They should look similar to the code below. Substitute your server name and SQL instance for each profile.

Data Source=[SERVER NAME]\[INSTANCE NAME];Integrated Security=True;Pooling=False

Important note here, once you switch the profile’s connection string to Windows Authentication, the Windows user account that you logged into your computer with, is the account that Visual Studio will now user to connect to the database. Make sure your Windows user account has at least the same level of permissions as the ‘aw_dev’ and ‘jenkins’ accounts. As a developer, you would likely have greater permissions than these two accounts.

Configuring Jenkins for Delivery of Script to Release

In many production environments, delivering or ‘turning over’ release-ready code to another team for deployment, as opposed to deploying the code directly, is common practice. A developer starts or ‘kicks off a build’ of the job in Jenkins, which generates artifact(s). Artifacts are usually logical collections of deployable code and other associated components and files, constituting the application being built. Artifacts are often separated by type, such as database, web, Windows services, web services, configuration files, and so forth. Each type may be deployed by a different team or to a different location. Our project will only have one artifact to deliver, the sql change script.

This first Jenkins job we create will just generate the change script, which will then be delivered to a specific remote location for later release. We start by creating what Jenkins refers to as a parameterized build job. It allows us to pass parameters to each build of our job. We pass the name of the configuration (same as our environment name) we want our build to target. With this single parameter, ‘TARGET_ENVIRONMENT’, we can use a single Jenkins job to target any environment we have configured by simply passing its name to the build; a very powerful, time-saving feature of Jenkins.

Step 1 - Parameterized Build Parameter

Step 1 – Parameterized Build Parameter

Let’s outline the steps we will configure our Jenkins job with, to deliver a change script for release:

  1. Copy the Solution from its current location to the Jenkins job’s workspace.
  2. Accept the target environment as a parameterized build parameter (ex. ‘Production’ or ‘Development’).
  3. Build the database project and its dependencies based on the environment parameter.
  4. Generate the sql change script based on the environment parameter.
  5. Compress and name the sql change script based on the environment parameter and build id.
  6. Deliver the compressed script artifact to a designated release location for deployment.
  7. Notify release team that the artifact is ready for release.
  8. Archive the build’s artifact(s).

Copy the Solution to Jenkins

I am not using a revision control system, such as TFS or Subversion, for our example. The Adventure Works Solution resides in a file directory, on my development machine. To copy the entire Solution from its current location into job’s workspace, we add a step in the Jenkins job to execute a simple xcopy command. With source control, you would replace the xcopy step with a similar step to retrieve the project from a specific branch)within the revision control system, using one of many Jenkins’ revision control plug-ins.

Step 2 - Copy Solution to Jenkins Workspace

Step 2 – Copy Solution to Jenkins Workspace

echo 'Copying Adventure Works Solution to Jenkins workspace...'
xcopy "[Path to your Project]\AdventureWorks2008" "%WORKSPACE%" /S /E /H /Y /R /EXCLUDE:[Path to exclude file]\[name of exclude file].txt

echo 'Deleting artifacts from previous builds...'
del "%WORKSPACE%\*" /F /Q

Excluding Solution files from Jenkins job’s workspace that are unnecessary for the job to succeed is good practice. Excluding files saves time during the xcopy and can make troubleshooting build problems easier. To exclude unneeded Solution files, use the xcopy command’s ‘exclude’ parameter. To use exclude, we must first create an exclude text file, listing the directories we don’t need copied, and call it using with the exclude parameter with the xcopy command. Make sure to change the path shown above to reflect the location and name of your exclude file. Here is a list of the directories I chose to exclude. They are either unused by the build, or created as part of the build, for example the sql directories and there subdirectories.


Build the Solution with Jenkins

Once the Solution’s files are copied into the Jenkins job’s workspace, we perform a build of the database project with an MSBuild build step, using the Jenkins MSBuild Plug-in. Jenkins executes the same MSBuild command Visual Studio would execute to build the project. Jenkins calls MSBuild, which in turn calls the MSBuild ‘Build’ target with parameters that specify the Solution configuration and platform to target.

Generate the Script with Jenkins

After Building the database project, in the same step as the build, we perform a publish of the database project. MSBuild calls the new SSDT’s ‘Publish’ target with parameters that specify the Solution configuration, target platform, publishing profile to use, and whether to only generate a sql change script, or publish the project’s changes directly to the database. In this first example, we are only generating a script. Note the use of the build parameter (%TARGET_ENVIRONMENT%) and environmental variables (%WORKSPACE%) in the MSBuild command. Again, a very powerful feature of Jenkins.

Step 3 - Build and Publish Project

Step 3 – Build and Publish Project


Compressing Artifacts with Apache Ant

To streamline the delivery, we will add a step to compress the change script using Jenkins Apache Ant Plug-in. Many consider Ant strictly a build tool for Java development. To the contrary, there are many tasks that can be automated for .NET developers with Ant. One particularly nice feature of Ant is its built-in support for zip compression.

Step 4 - Invoke Ant to Compress Artifact

Step 4 – Invoke Ant to Compress Artifact


The Ant plug-in calls Ant, which in turn calls an Ant buildfile, passing it the properties we give. First, create an Ant buildfile with a single task to zip the change script. To avoid confusion during release, Ant will also append the configuration name and unique Jenkins job build number to the filename. For example, ‘AdventureWorks.publish.sql’  becomes ‘’. This is accomplished by passing the configuration name (Jenkins parameterized build parameter) and the build number (Jenkins environmental variable), as parameters to the buildfile (shown above). The parameters, in the form of key-value-pairs, are treated as properties within the buildfile. Using Ant to zip and name the script literally took us one line of Ant code. The contents of the build.xml buildfile is shown below.

<?xml version="1.0" encoding="utf-8"?>
<project name="AdventureWorks2008" basedir="." default="default">
<description>SSDT Database Project Type ZIP Example</description>
<!-- Example configuration ant call with parameter:
ant -Dconfiguration=Development -DbuildNo=123 -->
<target name="default" description="ZIP sql deployment script">
<zip basedir="AdventureWorks2008/sql/${configuration}"
includes="*.publish.sql" />

Delivery of Artifacts

Lastly, we add a step to deliver the zipped script artifact to a ‘release’ location. Ideally, another team would retrieve and execute the change script against the database. Delivering the artifact to a remote location is easily accomplished using the Jenkins Artifact Deployer Plug-in. First, if it doesn’t already exist, create the location where you will deliver the scripts. Then, ensure Jenkins has permission to manage the location’s contents. In this example, the ‘release’ location is a shared folder I created. In order for Jenkins to access the ‘release’ location, give the ‘jenkins’ Windows user Read/Write (Change) permissions to the shared folder. With the deployment plug-in, you also have the option to delete the previous artifact(s) each time there is a new deployment, or leave them to accumulate.

Sharing Folder for Released Artifacts

Sharing Folder for Released Artifacts

Jenkins User Permissions for Shared Folder

Jenkins User Permissions for Shared Folder

Permissions for Shared Folder

Permissions for Shared Folder

Step 5 - Deploy Artifact to Release Location

Step 5 – Deploy Artifact to Release Location

Multiple Zipped Artifacts in Release Folder

Multiple Zipped Artifacts in Release Folder

Email Notification

Lastly, we want to alert the right team that artifacts have been turned-over for release. There are many plug-ins Jenkins to communicate with end-users or other system. We will use the Jenkins Email Extension Plug-in to email the release team. Configuring dynamic messages to include the parameterized build parameters and Jenkins’ environmental variables is easy with this plug-in. My sample message includes several variables in the body of the message, including target environment, target database, artifact name, and Jenkins build URL.

I had some trouble passing the Jenkins’ parameterized build parameter (‘TARGET_ENVIRONMENT’) to the email plug-in, until I found this post. The format required by the plug-in for the type of variable is a bit obscure as compared to Ant, MSBuild, or other plug-ins.

Step 6 - Email Notification

Artifact: AdventureWorks_${ENV,var="TARGET_ENVIRONMENT"}_${BUILD_NUMBER}
Environment: ${ENV,var="TARGET_ENVIRONMENT"}
Database: AdventureWorks
Jenkins Build URL: ${BUILD_URL}
Please contact Development for questions or problems regarding this release.
Release Request Notification Email Message

Release Request Notification Email Message

Publishing Directly to the Database

As the last demonstration in this series of posts, we will publish the project changes directly to the database. Good news, we have done 95% of the work already. We merely need to copy the Jenkins job we already created, change one step, remove three others steps, and we’re publishing! Start by creating a new Jenkins job by copying the existing script delivery job. Next, drop the Invoke Ant, Artifact Deployer, and Archive Artifacts steps from the job’s configuration. Lastly, set the last parameter of the MSBuild task, ‘UpdateDatabase’, to True from False. That’s it! Instead of creating the script, compressing it, and sending it to a location to be executed later, the changes are generated and applied to the database in a single step.

Hybrid Solution

If you are not comfortable with the direct approach, there is a middle ground between only generating a script and publishing directly to the database. You can keep a record of the changes made to the database as part of publishing. To do so, change the ‘UpdateDatabase’ parameter to True, and only drop the Artifact Deployer step; leave the Invoke Ant and Archive Artifacts steps. The resulting job generates the change script, publishes the changes to the database, and compresses and archives the script. You now have a record of the changes made to the database.


In this last of three posts we demonstrated the use of Jenkins and its plug-ins to created three jobs, representing three possible SSDT publishing workflows. Using the parameterized build feature of Jenkins, each job capable of being executed against any database environment that we have a configuration and publishing profile defined for. Hopefully, one of these three workflows may fit your particular release methodology.

Jenkins SSDT Jobs View

Jenkins SSDT Jobs View

, , , , , , , , , , , , , , , , , , , ,

  1. #1 by Peter Schott on August 8, 2012 - 3:59 pm

    You can also set up Jenkins to just do a build that produces the appropriate dacpac file and move that file up the chain through your processes using the SQLPackage.exe command to publish it. We’re doing that right now. Ideally, we’d like to get some “versioned” dacpac file ready to go so we can push that towards the upper environments and release just the versions that need to go as they’re ready.

    Just pointing out that you can generate a single dacpac and then use those publish profiles to push the changes to each server instead of generating a different script per server. Of course, if you need to customize or tweak the scripts, generating the script is definitely the better way to go.

    Thanks for the write-up.

    • #2 by Gary Stafford on August 8, 2012 - 5:04 pm

      We are currently using scripts because are DBAs are most comfortable with them. But, as we spin up new a few environments, I’d like to try the dacpac route you suggested. Thank you!

  1. Convert VS 2010 Database Project to SSDT and Automate Publishing with Jenkins – Part 1/3 « ProgrammaticPonderings
  2. Convert VS 2010 Database Project to SSDT and Automate Publishing with Jenkins – Part 2/3 « ProgrammaticPonderings
  3. Easy SMTP Server for Developers « ProgrammaticPonderings
  4. An architecture for automated team-based development and deployment of DWH/BI projects | Placko's SQL KB
  5. msbuild.exe is not recognized command after build in Jenkins | BlogoSfera

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: