Monday, November 30, 2015

Understand Persistence Context Collision in JEE

When deal with persistence layer in JEE application, persistence context is the magic that make your entity instances different from a normal POJO. Persistence context manages the entity instances to make it finally synchronized with database.

The persistence context collision happens when invoking stateful session bean from a stateless session method if both beans has their EntityManager defined. (BTW, Use stateful bean from stateless bean itself is not a good idea, but this article just focuses on persistence context)

In this article, for simplicity, when we mention stateless bean, we mean a stateless bean with EntityManager variable and its methods operate on the persistence. When  we mention stateful bean, we mean a stateful bean with EntityManager variable and its methods operate on persistence. In short we only talk about beans in persistence layer.

0. Basic rules about persistence context

Here are some basic rules about persistence context:

There will be only one active persistence context at any time for a transaction.

JEE container can propagate persistence context between different EntityManager variables in a single transaction. (Different EntityManager field variables of different beans can use the same persistence context).

Stateless bean usually use transaction-scoped persistence context, which means when the transaction is over, the persistence context is also gone. (When the bean's method is over, transaction is gone, persistence context is also gone)

Stateful bean usually use extended persistence context. The persistence is created when the bean instance is created and only destroy when the stateful bean is removed. The extended persistence context will be associated to a transaction when a method of stateful bean is called. When the method is over, transaction is gone, but the persistence context doesn't  go with the transaction but stay for the next transaction until the whole stateful bean is removed by the container.

Stateful bean always use his own extended  persistence context,  if the active persistence context is not the extended one, javax.ejb.EJBException will be thrown, this is called persistence context collision. This usually happens when call a stateful bean from a stateless bean.

Let see a simple example of persistence context collision.

1. Define pom.xml

<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/maven-v4_0_0.xsd">
<modelVersion>4.0.0</modelVersion>
<groupId>com.shengwang.demo</groupId>
<artifactId>jee-persistence-context-collision</artifactId>
<packaging>war</packaging>
<version>1.0</version>
<name>jee-persistence-context-collision Maven Webapp</name>
<url>http://maven.apache.org</url>

<properties>
<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
<!-- where glassfish 4.0 is installed -->
<glassfish.home>D:\glassfish4.0</glassfish.home>
</properties>

<dependencies>
<dependency>
<groupId>javax</groupId>
<artifactId>javaee-api</artifactId>
<version>7.0</version>
</dependency>
</dependencies>
<build>
<finalName>jee-persistence-context-collision</finalName>
<plugins>
<!-- Use Java 1.7 -->
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<version>2.5.1</version>
<configuration>
<source>1.7</source>
<target>1.7</target>
</configuration>
</plugin>

<!-- use mvn cargo:run to deploy and start server-->
<plugin>
<groupId>org.codehaus.cargo</groupId>
<artifactId>cargo-maven2-plugin</artifactId>
<inherited>true</inherited>
<configuration>
<container>
<containerId>glassfish4x</containerId>
<type>installed</type>
<home>${glassfish.home}</home>
</container>
<configuration>
<type>existing</type>
<home>${glassfish.home}/glassfish/domains</home>
</configuration>
</configuration>
</plugin>
</plugins>
</build>
</project>

The pom has 1 dependency for JEE 7 and 2 plugins. The first plugin specify the Java version (Java 1.7), the second one is for deploying package to local glassfish 4.0 with maven command-line. Demo use glassfish 4.0 as JEE container.


2. Define entity class


A very simple entity class Client.java with 2 fields, int clientId and String name.

package com.shengwang.demo.entity;

import javax.persistence.Column;
import javax.persistence.Entity;
import javax.persistence.GeneratedValue;
import javax.persistence.GenerationType;
import javax.persistence.Id;

@Entity
public class Client {
@Id
@GeneratedValue(strategy = GenerationType.IDENTITY)
@Column(name = "CLIENT_ID")
private int clientId;

private String name;

/* getter setter omitted */

@Override
public String toString() {
return "{" + clientId + "," + name + "}";
}
}

The entity is trivial.


3. Define stateful bean

package com.shengwang.demo.session;

import javax.ejb.Stateful;
import javax.persistence.EntityManager;
import javax.persistence.PersistenceContext;
import javax.persistence.PersistenceContextType;

import com.shengwang.demo.entity.Client;

@Stateful
public class MyStatefulBean {
@PersistenceContext (type=PersistenceContextType.EXTENDED) // extended
EntityManager em;
Client client;

public String changeClientName() {
if (client == null) {
client = em.find(Client.class, 1);
}
client.setName(client.getName() + "_" + "hello");

return client.toString();
}
}

This stateful bean just for demo, so doesn't make much sense. Its only method change the first client's name, add suffix to the name.


4. Define stateless bean

package com.shengwang.demo.session;

import javax.ejb.EJB;
import javax.ejb.Stateless;
import javax.persistence.EntityManager;
import javax.persistence.PersistenceContext;

import com.shengwang.demo.entity.Client;

@Stateless
public class MyStatelessBean {
@PersistenceContext
EntityManager em;

@EJB
MyStatefulBean statefulBean;

public String methodA() {
String c1 = em.find(Client.class, 1).toString(); // any operation of em
String c2 = statefulBean.changeClientName();

return c1 + "," + c2;
}
}

The stateless bean has one method, methodA. Call the em.find() first then call the stateful beans's method. Calling methodA() will cause persistence context collision!


Why? Because the transaction-scoped persistence is created lazy,  so it(PC-A) will be created only when em.find()  called, and this persistence context is now the active persistence context. But the stateful bean with extended persistence context init persistence context eagerly, which means the stateful bean already has a persistence context(PC-B) when it initialized. Now the active persistence context propagated to stateful bean,PC-A,   is not the extended persistence context PC-B. So collision happens and exception will be thrown. We can see this when we try to run it below.


5. Define a servlet as EJB client


To use the beans we defined above, let's define a simple servlet.

package com.shengwang.demo.servlet;

import java.io.IOException;
import java.io.PrintWriter;

import javax.ejb.EJB;
import javax.servlet.ServletException;
import javax.servlet.annotation.WebServlet;
import javax.servlet.http.HttpServlet;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;

import com.shengwang.demo.session.MyStatefulBean;
import com.shengwang.demo.session.MyStatelessBean;


@WebServlet(name = "testServlet", urlPatterns = { "/test" })
public class TestServlet extends HttpServlet {
private static final long serialVersionUID = 1L;

@EJB
MyStatelessBean statelessBean;

@EJB
MyStatefulBean statefulBean;

@Override
protected void doGet(HttpServletRequest req, HttpServletResponse resp) throws ServletException, IOException {
String str = statelessBean.methodA();

PrintWriter out = resp.getWriter();
out.printf("%s", str);
out.flush();
}
}

In the servlet, the stateless bean's only method is invoked.


6. Config persistence.xml


Usually the META-INFO/persistence.xml is simple for JEE applications, tell the server which data source application wants to use.

<?xml version="1.0" encoding="UTF-8"?>
<persistence version="2.1" xmlns="http://xmlns.jcp.org/xml/ns/persistence"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://xmlns.jcp.org/xml/ns/persistence
http://xmlns.jcp.org/xml/ns/persistence/persistence_2_1.xsd">
<persistence-unit name="demo-persistence-unit" transaction-type="JTA">
<jta-data-source>jdbc/MySQLDataSource</jta-data-source>
</persistence-unit>
</persistence>

There's a data source on server with JNDI name  jdbc/MySqlDataSource.


image


Finally, before running the demo, the project hierarchy looks like below.

image 

7. Run the demo


Run the demo with mvn command-line:

mvn clean verify cargo:run

This command will start glassfish server and deploy our application to it. Then access the servlet by browser. You can see the exception: javax.ejb.EJBException: There is an active transactional persistence context for the same EntityManagerFactory as the current stateful session bean's extended persistence context


image


8. What's more


What will happen if we make a little change to the stateless bean.

@Stateless
public class MyStatelessBean {
@PersistenceContext
EntityManager em;

@EJB
MyStatefulBean statefulBean;

public String methodA() {

// These 2 lines switch order
String c2 = statefulBean.changeClientName();
String c1 = em.find(Client.class, 1).toString();

return c1 + "," + c2;
}
}

The order of two lines are swithed. Now the demo can run without any exception.

Why? Because when invoke the stateful bean there is no persistence context yet, so the stateful bean's method can use its own extended persistence context to run. Then the em.find() is invoked, a transaction-scoped persistence create when first em operation invoked. So these 2 lines use different persistence context! If this is understood, then then output also makes sense to you.

image


Furthermore, If you really need to call stateful bean from stateless bean in it original order, which cause the persistence context collision. Here are some detours ( think twice when you want to do this):


1. Mark stateful bean (or its method invoked by stateless bean) REQUIRE_NEW for TransactionAttributeType. So there will be 2 transactions for stateless and stateful bean respectively. Each transaction can has its own persistence context. Similar scenario like above, entities may have different values on different context.


2. Mark stateful bean (or its method invoked by stateless bean) NOT_SUPPORT for TransactionAttributeType. If only invoke stateful bean's read-only operation from stateless bean.  (Servers like Glassfish may need to config JDBC connection pool to enable Non Transactional Connections)


3. Use Application-Managed EntityManager instead of the default Container-Managed Entity Manager in the stateful bean.

0 comments:

Post a Comment

Powered by Blogger.

About The Author

My Photo
Has been a senior software developer, project manager for 10+ years. Dedicate himself to Alcatel-Lucent and China Telecom for delivering software solutions.

Pages

Unordered List