This document outlines how Physical MDM customisations can be built from source artefacts in an automated build and test system. This document does not aim to be a complete guide on this topic, but rather to point the way to how some detailed steps can be implemented using examples.
The MDM Advanced or Standard editions both include the MDM Workbench. In version 11.0 and beyond the MDM Workbench is used by solution developers to create artefacts which customise the MDM solution for the physical, virtual and hybrid implementation styles. These source code artefacts are typically built into a Composite Bundle Archive (CBA) and deployed to WebSphere where they augment the functionality already available in the MDM Server Enterprise Business Application (EBA).
A good practice amongst MDM solution developers is to create an automated build process such that customisation source code is checked-in to a code control repository, and an automated build process takes those source files and builds the CBA ready for deploying onto post-build test systems, placing built artefacts into a second repository or shared file system.
Some automated systems take this “build” concept further, by automating the deployment of such built artefacts to test systems, which in turn report back on the “health” of the build, how many tests passed and failed, and generally quickly provide valuable feedback to developers whether recent changes broke the solution or not. Project managers overseeing such projects are able to reduce project risk by adopting this continuous delivery processes, and changes to MDM solutions become more reliable and safer as a result.
To add MDM solutions to such a continuous build environment it is necessary to:
This article is mostly concerned with step #7 – building source artefacts.
2. Materials and prerequisites
This article is accompanied by a collection of example scripts. We do not intend that these are used directly, but as an example of how you may wish to implement your own automated build process.
The current solution consists of four main files:
In order for the scripts to work, the machine running the scripts needs to have the following products installed:
To run the Ant scripts the user needs to run mdm_wb_build.xml as a build file.
The script contains only the “runBuild” target.
The target checks that necessary properties, such as Eclipse Home, date and time stamps and output folder prefix are set. Provided these properties do exist, it creates a folder based on OutputFolderPrefix and date and time, within which “logs”, “CBAExport” and “workspace” folders are created.
The logs folder contains “MDM
generateDevProject: BUILD SUCCESSFUL
workspaceBuild: BUILD SUCCESSFUL
exportCBA: BUILD SUCCESSFUL
End of report.
The CBAExport folder contains all of the exported CBAs.
The workspace folder contains a local copy of build artefacts.
After the directories have been created, the script checks which operating system it is running on and sets the isLinux or isWindows property to “true” as appropriate and calls either runAnt.sh or runAnt.bat to run a headless Eclipse process. The relevant file (either the batch or the shell script) should be available by default in the bin directory in the Eclipse installation directory.
The runAnt script then sets up the log files, environment variables and runs a second script “mdm
3. Step breakdown of the automated build and test system
Given that automated building and testing of MDM solutions is a worthwhile goal, the following sections provide some guidance in some of these areas where actions specific to the MDM tools and development/build environment are necessary, and some points of discussion are presented where choices exist.
3.1 Identify the pieces of the solution which represent the “source code” for the solution.
The source code for an MDM solution will be made up of a collection of Eclipse projects and their contents. MDM development, MDM configuration, MDM hybrid mapping, MDM service tailoring, MDM custom interface, MDM metadata and other MDM-specific projects types. CBA projects will add to the list.
MDM Development projects contain a “module.mdmxmi” file, which contains a model of the customizations which the project aims to create. This file should always be considered to be source code.
At some point the mdmxmi file will be used to generate Java, XML, SQL and other file artefacts, and there are a few different approaches you can take for these files:
The current solution is to only consider files which have been manually changed as “source code”, and “generate artefacts” from the mdmxmi model as part of the automated build process itself. This approach demands that the MDM workbench tools are installed as part of the build environment, because the “generate artefacts” process that turns .mdmxmi files into other artefacts will be a necessary part of the build process.
A project “MDM
3.2 Create a source code repository.
There are many choices regarding which product to use as a source code repository and covering them is not the aim of this document.
3.3 Recognize when a consistent set of code has been checked-in, at which point a “build” is started.
This event may be triggered manually, automated overnight, or whenever a change-set is delivered to the code stream. The capturing of this event is often specific to the code control system being used, though some solution teams augment this by adding a web page that enables build requests to be manually requested.
3.4 Create a build environment.
A build environment should include RAD (or RSA) which can be called in a “headless” manner such that functionality within RAD can be used without a user-interface being present.
MDM Workbench will be required in addition to RAD to perform a complete build of “module.mdmxmi” files.
For the list of platforms that MDM Workbench v11.0 and onwards support – refer to the product release documentation.
3.5. The build environment “boot-straps” itself.
A small script is responsible for “boot-strapping” the process by it checking-out the other build scripts which in turn build the artefacts from solution developers.
3.6 The build scripts check out the artefacts from code control to the local file system.
These actions are specific to the code control system so will not be discussed further here.
3.7. The source artefacts are processed, transforming them into built artefacts.
This phase of the automated system typically consists of a hierarchy of Ant files which decomposes the overall build process into many smaller steps and “Ant targets”. The Maven framework is a common choice of technology to oversee this phase.
These Ant files can be categorized into two types:
For a detailed walkthrough of specific implementations of the build process refer to the Ant scripts provided with this blog entry.
3.8. The build process often executes “unit tests” to further validate that the solution artefacts are healthy and do what they are expected to do.
The tools and approaches used to execute unit tests vary widely dependent on technology choice. Simple Java JUnit tests offer one simple solution, which can be invoked with scripts once the tests and tested code are built.
3.9 Built artefacts are published to a repository.
Every build against which build metadata can be gathered and reported is versioned by the publish process. Build logs, unit test results and results of other tests indicating the “health” of the build are gathered and published to the repository as well.
Products such as Rational Asset Manager can be used here, or for a really basic solution a simple shared folder on a network drive may suffice.
3.10 Automated deploy and test health of overall build.
If the build is considered “good” then further automation can be added to deploy the built solution to a test environment, with higher-level tests (functional and end-to-end system tests) exercising the solution further. Such tests can also report back to the build repository on the health of each build.
The automated deployment of entire systems for testing is often one of the most complex areas of the whole continuous development process. Products such as Rational Urban Code Deploy (UCD) can be used for this stage of the process, though for some environments a set of (reasonably complex) scripts might be sufficient.
For the MDM pieces, we are mostly concerned with deploying and un-deploying SQL scripts, depl
Prior to deploying extensions to the server, it is often necessary to modify the database. This is possible using the SQL scripts found in the MDMSharedResources project in the built workspace. Rollback scripts in the same location should be applied once testing is complete to reset the database back to a known good state.
For CBA deployment, Jython scripts can be used to manipulate the WebSphere server. Detailed documentation of these steps can be found in the WebSphere documentation.
Developing behavior extensions for InfoSphere MDM v11
Special thanks to Stephanie Hazlewood for providing guidance as well as content for some of the sections of this article!
Many established organizations end up having unmanaged master data. It may be the result of mergers and acquisitions or due to the independent maintenance of information repositories siloed by line of business (LOB) information. In either situation, the result is the same – useful information that could be shared and consistently maintained is not. Unmanaged master data leads to data inconsistency and inaccuracy.
One of the most fundamental extension mechanisms of InfoSphere MDM allows for the modification of service behavior. These extensions are commonly referred to as behavior extensions and the incredible flexibility they provide allows for an organization to implement their own “secret sauce” to the over 700 business services provided out of the box with InfoSphere MDM. The purpose of this tutorial is to introduce you to behavior extensions and guide you through the implementation, testing, packaging and deployment of these extensions. You will be introduced to the Open Service Gateway initiative (OSGi)-based extension approach introduced in the InfoSphere MDM Workbench as of version 11.
With the release of InfoSphere MDM v11, we adopt the OSGi specification which allows, amongst many other things, extensions to be deployed in a more flexible and modular way. This document will describe a real client behavior extension scenario and step you through all of the following, required steps:
- Extension scenario outline.
- Creation of the extension project.
- Development of the extension code.
- Deployment of the extension onto the MDM server.
- Testing deployed code using remote debugging.
We will then conclude this document with the summary of what you have learned.
It is often necessary to customize an MDM implementation in order to meet your solution requirements. One of the extension capabilities InfoSphere MDM provides it the ability to implement additional business rules or logic to a particular out-of-the-box service. These types of extensions are referred to as behavior extensions, as they ultimately change the behavior of a service. In this tutorial we will create a behavior extension to the searchPerson transaction.
The searchPerson transaction is used to retrieve information about a person when provided with a set of search criteria. You can filter out the result set by active, inactive or all records that get retrieved by these criteria. Important to note is that this particular search transaction uses exact match and wildcard characters to retrieve the result set. There are separate APIs available for probabilistic searching – this service is not one of them.
Sometimes, the searchPerson transaction response may contain duplicate parties. For example, if a party contains both legal and business names which are identical, and searchPerson transaction uses last name as criteria, - the parent object will be returned twice in the response, as it will be matched by both of the names. While this behavior is acceptable in some circumstances, some cases might more filtering before it is returned. In order to do so, we will create a behavior extension, which will be responsible for processing transaction output and removing any duplicate records in the result set. The InfoSphere MDM Workbench provides us with exactly the right tools to quickly create and deploy such an extension.
Creating extension project
First, create the extension project structure using the wizards provided by MDM Workbench. Go to File -> New -> Other… and search for Development Project wizard:
If you cannot find Development Project wizard within the list, chances are the Workbench has not been installed, please verify using IBM Installation Manager.
When creating your project, make sure to specify a unique project and package names in order to avoid conflict with the existing ones:
Make sure to choose the correct server runtime for your projects, as well as unique name for the CBA project:
Note: You are allowed to choose from the existing CBAs. A single CBA can contain multiple development project bundles.
Click Finish and wait for the wizard to generate the required assets.
At this point, what we have is a skeletal InfoSphere MDM Development project that contains all of the basic facilities to help us create the desired extension. The next step is to create the extension assets and there are two ways of doing so: either by using the behavior extension wizard, or by using the model editor.
Creating a behavior extension using the extension wizard
You can create an extension using a wizard in the MDM Workbench, much like the one used to create a development project:
1. Open Behavior Extension wizard by going to File -> New -> Other… -> Behavior Extension, located under Master Data Management -> Extension folders
Note: A development projects can contain multiple extensions of various types underneath it. You might choose to use development projects to logically group extensions having a similar purpose, type or to facilitate parallel development activities.
3. Within the next window, choose a name and a description for your behavior extension. Choose a Java class name for your extension. This is the class that we will be populating with custom logic in order to achieve desired behavior. Alternatively, if you require to use an IBM Operational Decision Manager (ODM, and previously known as ILOG) rule – specify this associated parameter. ILOG/ODM rule creation is not covered as a part of this tutorial as we will implement the extension in as a Java extension.
4. Within the “Specify details of the trigger” pane, you need to specify the following parameters:
a. Trigger type:
i. ‘Action‘ will cause the behavior extension to trigger whenever chosen transaction is ran by itself or as a part of another transaction. ‘Actions’ are executed at the component level. .
ii. On the other hand, if you are looking to trigger extension only on a specific standalone transaction event (otherwise known as controller level transaction) select ‘Transaction’ trigger type.
iii. ‘Action Category’ trigger type executes behavior extension on various data actions such as add, update, view or all for extensions to be executed at the component level.
iv. ‘Transaction Category’ trigger type will kick off behavior extension when a transaction of specified type is executed, namely inquiry, persistence or all.
b. When to trigger:
i. ‘Trigger before’ will cause the behavior extension to fire of before the work of the transaction is carried out. Sometimes you will hear this referred to as a preExecute extension. It is a typically used when some sort of preparation procedure has to be executed before the rest of the transaction is carried out. An example of such scenario would be preparing data within the business object that is being persisted.
ii. ‘Trigger after’ will cause the behavior extension to run after the transaction work has carried out. Sometimes you will hear this referred to as a postExecute extension. It is typically used in the scenarios where logic implemented in the behavior extension depends on the result of the transaction. Normally any sort of asynchronous notification would be placed in a post behavior extension, as there would be no way to roll it back in case of transaction failure, if it is sent before the transaction is executed.
c. ‘Priority’ parameter indicates the order in which this behavior extension will be triggered. The lower the priority number, the higher the priority. That is, a behavior extension with priority 1 would execute first followed by behavior extension with priority 2, 3 or 4 in that order.
In our scenario we are looking to filter the response of a specific transaction,, namely searchPerson. Therefore we set the trigger type to be ‘Transaction’ with value of searchPerson. Since we are filtering the response of the transaction – we have to trigger our behavior extension after the transaction has gone through, and response became available. Lastly, in our particular example priority does not play a special role, so we will leave it at default of ‘1’.
5. After the above configuration is done, click Next and review the chosen parameters. Note that there is a checkbox at the top of the dialog, allowing you to generate the code based on the specified parameters immediately. For the purposes of this tutorial leave it checked and click Finish. The workbench will generate all of the required assets for you.
Creating a behavior extension using the model editor
If you have used the wizard approach above to create the behavior extension already, please feel free to skip ahead to the section titled, “review generated assets” that follows.
This section describes how to generate a behavior extension using the model editor. To do so, the following steps will guide you through this process:
1. Go to the development project you created earlier and open the module.mdmxmi file under the root folder of the project. Select the model tab within the opened view.
2. Right click Part
4. Now we will create a transaction event definition under behavior extension. Right-click the behavior extension, then select New - > Transaction Event.
5. Once the transaction even has been created, specify the appropriate properties:
a. Because this event is triggered on the personSearch transaction, PersonSearchEvent is appropriate. Recall that sometimes the “trigger before” behavior is referred to as “preExecute” extension.
b. Because ‘Pre’ checkbox stands for preExecute, (more specifically the behavior extension gets executed before the rest of the transaction) leave it unchecked. Similar to the wizard configuration, leave priority as ‘1’, since priority of execution does not affect this behavior extension.
c. Finally, select searchPerson as the transaction of choice by clicking Edit… -> Party -> CoreParty -> searchPerson.
After all of the above configurations are done and reviewed, go ahead and click Generate Code under the Model Actions section of the view, telling workbench to generate configured assets.
Review your generated extension code
Once either of the above methods is used, let us review the generated assets:
o EXTENSIONSET table record defines the behavior extension, its associated class best
o CDCONDITIONVALTP defines a new condition of transaction name being equal to searchPerson.
o EXTSETCONDVAL connects the above CDCONDITIONVALTP record to the behavior extension record from EXTENSIONSET. Additionally another EXTSETCONDVAL record connects CDCONDITIONVALTP with id of ‘9’, which stands for execution of behavior transaction after transaction.
Let us now move on to developing the extension code required to filter out duplicate person records from the result set returned by the searchPerson transaction.
Develop the extension code
The behavior extension skeleton and supporting configuration assets have now been generated. You add your custom logic, or behavior change, in the execute method of Pers
public void exec
// Only work with vectors in the response
// Get the response object hierarchy
// Iterate through the party search result
// objects to find duplicates
Iterator listIterator =
// We will keep the party ids of objects we've already
// processed to identify the duplicates
Vector partyIdList = new Vector();
Object o = list
if(o instanceof TCRM
String partyId = pers
// If the party id has not been seen yet, this person
// object is not a duplicate, otherwise - remove it from
// the response
Note: The above implementation is not pagination friendly and pagination will not be covered as a part of this tutorial.
Once you have compiled the code above, you will notice that some of the classes are not found and have to be imported. You cannot simply import TCRM
After recompiling the projects again, you will notice that the Part
This error is occurring because the composite bundle that contains Part
Now that all compilation problems have been resolved, we are ready to deploy our extension onto the server.
Deploying your new behavior extension to MDM
Once the implementation of the behavior extension has been developed, we are ready to deploy it onto the server. There are two steps involved into the deployment:
- Deploying code to the server.
- Executing generated SQLs to insert required metadata.
Deploying code to the server
Our customized behavior extension can be deployed to the server as a Composite Bundle Archive (CBA) as follows:
1. Make sure that the customized code has been built and then export the CBA containing the behavior extension by right clicking the CBA project and selecting ‘Export… -> OSGi Composite Bundle (CBA)’.
2. In the opened view, select Part
3. Click ‘Finish’. The CBA containing the behavior extension has now been exported to selected location.
4. At this point, we will assume that the MDM instance is up and running. Let’s open the WebSphere Administrative Console. We are looking to import our new CBA into the internal bundle repository. To do so go to Environment -> OSGi bundle repositories -> Internal bundle repository. In the opened view, click New…, choose Local file system and specify the location of the CBA we’ve exported above. Save your progress.
5. Once the CBA has been imported, attach this new bundle to the MDM application. Go to Applications -> Applications Types -> Business-level applications. Choose MDM application from the opened view. In the next open view, open the MDM .eba file.
6. We are now looking at the properties of the MDM Enterprise Bundle Archive (EBA). In order to attach our CBA, go to Additional Properties section and select Extensions for this composition unit.
7. If this is the first extension that you’ve deployed on your instance, the list of attached extensions will be empty. Let’s now click Add…, and check the CBA we’ve imported above, then click Add. Wait for the addition to complete. Save your changes.
8. You may think that we are done here, but not quite. We’ve only updated the definition of the EBA deployment by adding our extension. The MDM OSGi application itself has not been updated and even if you restart the server, your new behavior extension will not be picked up. So you must update the MDM application to the latest deployment by returning to the EBA properties view.
Before we attached our extension, the button shown above was grayed out; the comment stated that the application is up to date. But since we’ve update our application with a new extension bundle, we need to update it to the latest deployment. Go ahead and click the Update to latest deployment … button.
9. In the next view, you can see that the Part
At this point, scroll down and click Ok to proceed. It may take several minutes depending on your system hardware.
10. At this point, WebSphere will take you through three views, offering multiple information summaries of the deployments and several customization options. There is no need to customize anything, go ahead and click Next three times, followed by Finish. At this point the application will update. It may take some time; please allow 5 – 10 minutes to complete depending on underlying hardware. Once it is complete – save your changes. At this point, the MDM application has been updated to the latest deployment which includes our extension.
Now we need to deploy our custom metadata to the database. This metadata will govern the behavior of our extension in ways discussed above.
Deploy metadata onto the MDM database
As mentioned earlier in this tutorial, the Workbench generates database scripts that insert the required configuration into the metadata tables of the MDM repository. This metadata is generated based on the parameters we provided for our behavior extension as part of the Creating extension project section. In order to deploy this metadata to the database, run the database scripts listed under the resources -> sql folder that are appropriate for your database type. Conversely, if you need to remove extension from the server, you would need to run the rollback scripts provided in the same folder.
Note: In the case where some portion of the script fails, please investigate the error, because it may render the extension useless. Potential reasons for an error may include residual data from previous extension (rollback was not run when extension was removed), incorrect database schema, etc.
Once the scripts have been successfully run, you’re your behavior extension has now been successfully deployed. Restart your WebSphere server so that your new metadata gets picked up when the application runs next.
Testing deployed code using remote debugging
Now that all of the aspects of the behavior extensions have been deployed, we are ready to test it out! To do that, run a searchPerson transaction. It is required to have at least one person in the database so that you can actually search and yield a successful search result to trigger your new extension. This test will show us that the extension is successfully deployed. Once the transaction returns as successful, go to the SystemOut.log of the WebSphere server which is located under the log folder of the WebSphere profile where MDM application is deployed. If the extension has deployed correctly, due to the following line in our custom code:
You should be able to see this message in the logs:
[6/17/14 13:24:59:816 EDT] 000001b3 SystemOut O Part
Note: The log message is there for testing purposes only, and depending on the usage of the behavior extension can significantly impede performance. For that reason please make sure to remove such debugging messages or put them into fine logging level before going into production. Such as:
Configuring WebSphere Application Server debug mode
To observe the behavior of our extension more closely, put WebSphere server into the debug mode, and connect MDM Workbench to the said server in order to debug our code step by step. To put your server in debug mode:
1. Go to WebSphere Application Server administrative console, and navigate to Servers -> Server Types -> WebSphere application server -> <Name of your instance>.
2. Once in the server configuration view, take a look at Server Infrastructure section and navigate to Java and Process Management -> Process definition.
3. In the Additional Properties section, select Java Virtual Machine.
4. Once we are in the Java Virtual Machine view, navigate down to the Debug Mode checkbox, check it and provide the following settings in the Debug arguments textbox:
Note that ‘7777’ is the debug port to which the MDM Workbench will connect. Make sure this port does not conflict with any other assigned ports on the server, and set it accordingly.
5. Save configuration and restart your server. It is now running in debug mode. Note: If later you observe unexpected performance degradation and do not require debug mode any longer, make sure to take the server out of the debug mode using the same steps.
Configuring MDM Workbench to for remote debugging
Once the server is running in debug mode, we can go back to the MDM Workbench and configure it for debugging:
1. In MDM Workbench, go to Run -> Debug Configurations.
2. Within the Debug Configurations window, double click Remote Java Application. This will create a new Remote Java Application profile.
3. When configuring the Remote Java Application, lets name our configuration ‘MDM Local Instance Debug’. The Project setting does not play a role, you may leave it empty, or whatever the default populated value is. Connection Type should remain as ‘Standard (Socket Attach)’. Lastly Connection Properties should reflect the location of the MDM instance and debug port we’ve chosen above.
We will not cover other tabs because the configuration we’ve done so far is sufficient.
4. Once configuration is complete, hit Apply followed by Debug in order to attach to the MDM instance. The attach process may take a little bit of time depending on the environment. Once it is complete, go to the Debug perspective of your environment. In the debug view, you should observe the connected MDM instance if the attach was successful:
You can see above that the instance is available along with all of the threads.
5. Finally set a break point at the beginning of the behavior extension execute method and observe this breakpoint getting engaged once we run a searchPerson transaction:
6. If you have multiple TCRM
As a last point, note that we can debug both local and remote instances as described above, using Eclipse’s Remote Java Application debug capabilities.
In this tutorial we’ve gone through the steps of creating, configuring, deploying and testing a basic yet realistic behavior extension scenario for InfoSphere MDM.
We’ve covered two ways in which an extension template can be created: while the wizard option is straightforward and is preferable for a novice or a simple extension scenario, the model editor allows for more flexibility.
We’ve taken a look at the various configurations that apply to a behavior extension and outlined their effects on its execution. Additionally, we’ve covered the assets that get generated as a result of the configuration.
For the development step, we’ve created and analyzed the implementation of our behavior extension.
And finally, we’ve deployed, tested and debugged our behavior extension to make sure it performs as expected.
All of the above steps constitute a complete development process of an MDM Server behavior extension.
As of version 10.1, MDM is secured by default. This means that using the Web Service Explorer to test your web service will not be possible. Whilst there are many web service testing tools out there like SOAPUI, there is one that is included within the MDM Workbench that you can use, the Generic Service Client. The following steps detail how to invoke the required MDM web service :
This will then open the following editor :
The responses are saved and can be rerun, but if you want more functionality you'll probably need to look at Rational Performance Tester
InfoSphere Master Data Management operational server v11.x OSGi best practices and troubleshooting guide
Dany Drouin 270004VXKT Visits (5035)
InfoSphere Master Data Management operational server v11.x OSGi best practices and troubleshooting guide
Note: This preview only covers the initial set up of the MDM Workbench. The
The goal of this article is to show best practices and optimal development practices for developing with the InfoSphere MDM operational server. We will discuss common OSGi patterns, troubleshooting, including failures and resolution, as well as how to best deploy MDM composite bundle (CBA) extensions.
The InfoSphere MDM version v11 operational server is based on an enterprise OSGi architecture, which is modular in nature. The benefits of a modular architecture application design include reducing complexity, reducing time to delivery, and increasing serviceability. The Java EE infrastructure leveraged in previous versions of InfoSphere MDM had limited ability to enforce or encourage a modular design.
The advantage of a modular MDM application is to allow customization to be deployed without having to alter the core MDM application. Instead, customizations are attached in the form of extensions to the core MDM application. This is done using composite bundles, or CBA files.
Optimal workspace operational server configurations
The InfoSphere MDM Workbench is a tool that supports development of customizations and extensions to MDM operational server. The MDM Workbench allows you to define the desired data model and transactions and then generates the code required to implement the MDM Server extensions.
When using workbench to build and deploy your MDM customizations and extensions, there are a few workspace configurations to consider for achieving the best performance and development experience.
This blog post details how to use the template models provided in MDM Workbench to create and deploy a new Virtual Configuration Project to an Operational Server, and then deploy the sample data supplied in the template model.
Creating a new Virtual Configuration Project
Creating a connection to the Operational Server
Deploy the new Configuration Project:
Processing and loading the sample data: