EJB 3.x module packaging overview

EJB modules without deployment descriptors

You can package EJB modules without a deployment descriptor if you are using EJB 3.x beans. To do this, you must create a JAR file or WAR file with metadata in an annotation which is located in the EJB component. EJB 3.x beans do not need an entry in the ejb-jar.xml file for metadata that you have defined through annotations.

With EJB 3.0, the default was to scan annotations during the installation of an EJB 3.0 module. For WebSphere® Application Server, Version 8.5, the default is not to scan pre-Java EE 5 modules during the application installation or at server startup.

To preserve backward compatibility with both the Feature Pack for EJB 3.0 and the Feature Pack for Web Services, you have a choice whether to scan legacy web modules for additional metadata. A server level switch is defined for each feature pack scan behavior. If the default is not appropriate, the switch must be set on each server and administrative server that requires a change in the default. The switches are server custom properties com.ibm.websphere.webservices.UseWSFEP61ScanPolicy={true|false} and com.ibm.websphere.ejb.UseEJB61FEPScanPolicy={true|false}. To define these properties in the administrative console click Application servers > server name > Process definition > Java Virtual Machine > Custom properties.

EJB modules with deployment descriptors

You can continue to use EJB modules with deployment descriptors. Modules with deployment descriptors can support any EJB specification version level, including EJB 3.x, but generally these descriptors should reflect the implementation requirements of the components in the module.

An EJB module can have an EJB 3.x, 2.x, or 1.x deployment descriptor.

For EJB 2.x or 1.x deployment descriptors, it is assumed that the deployment descriptor contains the full metadata for the module, and no additional scanning of annotation metadata occurs.

The EJB container annotation scanning is performed on EJB modules that either have no deployment descriptor or have an ejb-jar.xml deployment descriptor at the EJB 3.0 schema level with the metadata-complete XML attribute set to false or omitted. See the Annotation scanning behavior section for the complete set of rules used by the server to determine if annotation scanning is performed.

Annotation scanning behavior

The server can inspect the class files in the module for annotation content. The server searches for annotation content that might define a component, a reference to a resource, or a particular behavior. For example, annotations might be used to define an EJB component, or to declare a reference to a data source that must be used by an EJB component, or to declare the transactional or security attributes that are associated with an EJB method. This inspection process is referred to as annotation scanning. If the class files in the module contain annotations that must be respected by the server, then the server must be configured so that annotation scanning occurs. If the class files in the module do not contain annotations, then for performance reasons you can configure the server so that annotation scanning does not occur.

The following tables indicate how the decision to scan, or not scan, annotations is made for EJB content that is packaged in an EJB JAR module or a WAR module.

Persistence units

Persistence units, including the persistence.xml file and the classes associated with it, can be packaged in the module for which they are required. They can also be packaged in the separate utility JAR file that is packaged in the EAR file with its dependent module.

Application packaging

You can mix EJB 2.x and earlier beans with EJB 3.x beans in the same application. However, EJB 3.x beans are not recognized in EJB 2.x or EJB 1.x modules.

AutoLink

JPA packaging

It is recommend that persistence units be packaged in separate JAR files to make them more accessible and reusable. These can be tested outside the container, with or without actual database persistence occurring. Persistence units can be included in stand-alone applications or into EAR files as utility JAR files. Because of the variety of use cases and potential performance issues when scanning large quantities of classes, it is recommended that the persistence unit defines the classes of the persistence units.

Session facades used for persistence scenario

A common pattern is to use session facades for persistence. Using session bean facades to drive JPA is supported. The EntityManager interface is not thread safe, therefore, servlets should never inject @PersistenceContext. Servlets must either use the facade pattern or use an EntityManagerFactory instance to create an EntityManager on each request.

It is recommended that JPA persistence units be defined in a separate JAR file, apart from the session bean facades. Not only is this a best practice that gives greater flexibility in sharing, it also avoids problems mixing JPA and non-JPA annotated classes.

Typically, a JAR file is created to hold the entity classes and the JPA persistence.xml definition and added to the EAR file as a utility JAR file. The EJB 3.x module adds a dependency on the JAR file by declaring it in the EJB 3.x module MANIFEST.MF. For example, if an EAR contains a TradeApp.ear, TradeWeb.war, EJB3Trade.jar, and TradeInfo.jar file, the EJB3Trade.jar file would have a MANIFEST.MF that looks like the following:

Manifest-Version: 1.0
Class-Path: TradeInfo.jar

The session facade in the EJB3Trade.jar file refers to JPA entity classes and persistence units in the TradeInfo.jar file. The web application defined in the TradeWeb.war file can do the same to work with the JPA entity objects as Data Transfer Objects flowing between the web and EJB container tiers.

Cross-tier and cross version session bean reference scenario

There are several ways to define and use references to EJB 3.x session beans. For EJB 3.x session to session, the @EJB injection target can be used. For cross-tier, for example, web application to EJB 3.x session, or cross-version, for example, EJB 2.1 session to EJB 3.x session, an XML deployment descriptor reference can be used to define ejb-refs and ejb-local-refs. There are two variations of these, depending on whether an EJB 3.x business interface is referred to, or a pre-EJB 3.x component-style interface that also defines an EJBLocalHome is referred to. Web applications and client applications can also use the @EJB annotation if the component being referenced can be resolved using autolink.

The previous scenario uses an EJB 2.1-style client pattern with an EJB 3.x-style session bean implementation. For a more current client style, the client-side can be cleaned up to look up the session bean business interface directly, rather than going through a home interface. In this case, it is not necessary to define the @LocalHome(<localhome>.class) annotation. You can use a variant definition of ejb-ref and ejb-local-ref to do this. Use a null value for the local-home element value and bind the ejb-local-ref to the session bean's ejblocal: binding rather than the home binding. For example:

<ejb-local-ref id="EJBLocalRef_1154112538064">
	<description>com.ibm.persistence.ejb3.order.facadecom.ibm.persistence.ejb3.order.facade</description>
	<ejb-ref-name>ejb/OrderEJB</ejb-ref-name>
	<ejb-ref-type>Session</ejb-ref-type>
	<local-home></local-home>
	<local>com.ibm.persistence.ejb3.order.facade.OrderProcessor</local>
</ejb-local-ref>

The client code also must be adjusted to do the appropriate casting for the object being looked up. In this case, the business interface instead of the home interface:

try {
	InitialContext ctx = new InitialContext();
	orderProcessor = (OrderProcessor)ctx.lookup("java:comp/env/ejb/OrderEJB");
}
catch(Exception e)  {
	e.printStackTrace(System.out);
	throw new ServletException(e);
}