24 Oracle Complex Event Processing

24.1.1 Deprecated API and Schemas

Table 24-1 lists Java API deprecated in 11g Release 1 (11.1.1) and the replacement API, if any.

Table 24-2 lists Oracle CEP schema deprecated in 11g Release 1 (11.1.1) and the replacement schema, if any.

For information on deprecated Oracle CQL syntax, see:

• Section 24.3.9, "Outer Join Operator "(+)" is Deprecated"

24.1.2 Adapter Changes

The programming model for adapters has changed as follows:

• Outbound Adapters

• Lifecycle Callback Methods for Adapters

• Runnable Adapters

• Adapter Factories

• Publishing Adapter Providers

• StockTick Event Type Replaced by OracleStockTick

Outbound Adapters

Adapters can now also be outbound; previously they were only inbound.

Lifecycle Callback Methods for Adapters

Spring tags for adapters now include attributes for setting lifecycle callback methods via the EPN assembly file.

Runnable Adapters

For adapters that want to run in a thread, their Java class should now implement com.bea.wlevs.ede.api.RunnableBean.

Adapter Factories

Programmers are no longer required to create an adapter factory when creating adapters. You only need to create an adapter factory if you want to share adapters among applications.

Publishing Adapter Providers

Adapter providers are factories for adapters that are published through the OSGi service registry.

Previously, the only way of advertising adapter providers was to use Spring-DM only, like this:

<osgi:service interface="com.bea.wlevs.ede.api.AdapterFactory" ref="myBean">
    <osgi:service-properties>
        <entry key="type" value="SocketAdapterType"/>
    </osgi:service-properties>
</osgi:service>

However, Spring-DM does not support exporting and importing an OSGi service in the same application so the above approach is deprecated.

Now, you should declare providers with the wlevs:factory tag like this:

<wlevs:factory provider-name="SocketAdapterType" ref="myBean">

This will work whether the adapters are in the same application or not.

Note that if your adapters and providers are collocated (both in the same application) using a provider is optional; you could just instantiate the adapter directly.

Failure to make this change may result in an application that fails to start and eventually times out.

StockTick Event Type Replaced by OracleStockTick

When using the loadgen adapter, note that the StockTick event is now named OracleStockTick.

24.1.3 New and Deprecated Options in the Deployer Tool

The -start and -stop commands of the Deployer tool have been deprecated.

When using the -install command to install an application, Oracle CEP automatically starts it after all internal initialization tasks have completed. Subsequently, if you stop and start the Oracle CEP server instance, the application is automatically stopped and started, respectively.

The Deployer tool has two new options: -suspend and -resume. Users should use -suspend to suspend a currently running application, and -resume for it to resume running.

"Deployer Command-Line Reference" in the Oracle CEP Administrator's Guide.

24.1.4 Changes in Management, Monitoring, and JMX

The management framework of Oracle CEP has been overhauled in this release.

Oracle CEP no longer supports the JRMP protocol. Instead, JMX clients must use the more secure MSA protocol for both local and remote access to the Oracle CEP JMX server. When you connect to the Oracle CEP JMX server that is running on localhost or on a remote host, you must always use the JMX URL service:jmx:msarmi://HOST-NAME:PORT/jndi/jmxconnector so that you are always using the MSA connector (where HOST-NAME is either localhost or the name of the remote host and PORT is the Oracle CEP server JNDI port).

For more information, see:

• "Accessing the Oracle CEP JMX Server" in the Oracle CEP Administrator's Guide

• "Configuring JMX" in the Oracle CEP Administrator's Guide

24.1.5 Using stopwlevs.sh With Multiple Servers on the Same Host

Consider a scenario where you start two servers on the same host, for example, with server 1 listening on port 9002 and server 2 listening on port 9022. If you then use stopwlevs.sh in the server 1 domain directory, the server listening on port 9002 is stopped. If you then use stopwlevs.sh in the server 2 domain directory, the stop operation fails due to connection reset error.

Workaround: use the stopwlevs.sh command line argument -url or -listenPort to specify a port other than 9002.

24.1.6 Monitor Service may Underestimate Latency For Application-Timestamped Channels

The monitor service may underestimate the latency time of events that are being processed through the CQL processor when application-timestamped channels are being used and is-totally-ordered is not enabled.

24.1.7 Connecting to Oracle CEP Server Using JConsole may Throw Benign Null Pointer Exceptions

When you connect to Oracle CEP server using JConsole (with or without the wlevsjconsole.sh script), JConsole may throw a java.lang.NullPointerException.

Workaround: click OK in the exception dialog box and proceed. The exception does not affect the JMX connection or Oracle CEP server.

24.1.8 Variable Duration Non-Event Detection is not Supported in This Release

Fixed duration non-event pattern detection is supported in 11g Release 1 (11.1.1). When you create a query for fixed duration non-event detection, you may use the DURATION clause with constant value and time unit, such as DURATION 5 SECONDS, or just a constant value such as DURATION 5.

Variable duration non-event pattern detection is not supported in 11g Release 1 (11.1.1). That is, you may not use the DURATION clause with an arbitrary arithmetic expression, such as DURATION c1+4.

Recurring non-event pattern detection is supported in 11g Release 1 (11.1.1) but only for the fixed duration case. That is, you may use a DURATION clause with a MULTIPLES OF clause but only for a duration that is a constant value.

For documentation errata related to this issue, see Section 24.3.7, "Variable Duration Non-Event Detection is not Supported in This Release".

24.1.9 JMSAdapterMBean Getter Methods for New Connection and User and Password

The following methods have been added to com.bea.wlevs.management.configuration.JMSAdapterMBean:

• getConnectionUser

• getConnectionPassword

• getConnectionEncryptedPassword

The following attributes have been added to the wlevs_application_config.xsd element jms-adapter:

• connection-user

• connection-password

• connection-encrypted-password

When Oracle CEP acquires the JNDI InitialContext, it uses the user and password (or encrypted-password) settings.

When Oracle CEP calls the createConnection method on the javax.jms.ConnectionFactory to create a connection to the JMS destination (JMS queue or topic), it uses the connection-user and connection-password (or connection-encrypted-password) settings, if configured. Otherwise, Oracle CEP uses the user and password (or encrypted-password) settings.

You can use the connection-user and connection-password (or connection-encrypted-password) settings in applications where one security provider is used for JNDI access and a separate security provider is used for JMS access.

24.2.1 Accessing Signal Generation Dashboard With Firefox Version 3.0 on Windows

When accessing the signal generation example dashboard (at URL http://localhost:9002/signalgeneration/dashboard.html) using the Firefox browser version 3.0 on Windows, you will get an error when you click Start.

Workaround: use Internet Explorer version 7 to access the signal generation example dashboard.

24.2.2 Starting a Server in a Domain Without Deployed Applications

When you start a server in a domain with no deployed applications, the Server STARTED message may not appear in the command window from which you ran the server start script, even though the server has successfully started.

24.2.3 VPN Software and Multicast Traffic

Active VPN software is known to have unpredictable behavior on multicast traffic. Additionally, having both Cisco VPN and Nortel VPN installed breaks multicast traffic.

24.2.4 Starting Oracle CEP Visualizer With Firefox

When using Oracle CEP Visualizer with the Firefox browser (version 3.0, SSL, and JDK 1.6), the first time you access a Oracle CEP Visualizer page it will render slowly, sometimes up to 30 seconds.

24.2.5 Uploading Files Using Oracle CEP Visualizer With Firefox and SSL

When using Oracle CEP Visualizer, you may get an error if you try to upload a file using the Firefox browser and SSL.

This is a known problem (http://bugs.adobe.com/jira/browse/FP-226) with Adobe Flex that affects Oracle CEP Visualizer.

24.2.6 A Channel Cannot be a Foreign Stage When Connected to an Oracle CQL Processor

If a channel is connected to an Oracle CQL processor, it cannot be a foreign stage.

24.2.7 Exceptions Thrown When Using Oracle CEP Visualizer with FireFox

When using Oracle CEP Visualizer with FireFox:

• Do not set the Jetty scratch directory to a path with a space in it. Doing so will cause FileNotFoundException.

• Ignore EOFException thrown on the server side; these exceptions will not cause client side failure.

24.2.8 Oracle CEP IDE for Eclipse is Incompatible With Required FireFox 3 Libraries on Linux

The Oracle CEP IDE for Eclipse uses Eclipse 3.3, which is incompatible with required FireFox 3 libraries on Linux. In this case, Eclipse will crash on startup with a JVM core dump and error message like this:

# An unexpected error has been detected by Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00bc80ac, pid=18044, tid=3086637584
#
# Java VM: Java HotSpot(TM) Client VM (11.3-b02 mixed mode, sharing
linux-x86)
# Problematic frame:
# C  [libc.so.6+0x710ac]  memcpy+0x1c
#
# An error report file with more information is saved as:
# /usr/local/bt/workspace/testruntime/eclipse/hs_err_pid18044.log 

Workaround: perform the following procedure.

1. Change to the root user on your Linux machine.

2. Prevent Eclipse from loading the libc.so library noted in the core dump error message:

   mv libc.so libc.so.orig
   

   Note:

   This will disable FireFox 3 on this machine. To use FireFox 3, you must change the name of this library back to libc.so.

3. Follow the instructions outlined in the Eclipse Linux crash support Wiki page:

   http://wiki.eclipse.org/IRC_FAQ#I_just_installed_Eclipse_on_Linux.2C_but_it_does_not_start._What_is_the_problem.3F

   In particular, ensure that the JVM you are using with Eclipse is:

   • explicitly pointed to in the eclipse.ini file, and

   • JDK 1.6 or higher

24.2.9 Configuring Oracle CEP Native Clustering Using IPv6 on Linux

When using Oracle CEP native clustering on Linux with IPv6, you may see an error like the one shown in Example 24-1 after restarting a member of a cluster.

<Error> <evs4j> <BEA-2049005> <The cluster member 3 is considered to have failed>

Workaround: perform the following procedure:

1. Using the editor of your choice, open the startwlevs.sh script for the affected Oracle CEP sever.

   The startwlevs.sh server start script is located in the server directory under the main domain directory. For example, the default server directory of the HelloWorld domain is located in ORACLE_CEP_HOME/ocep_11.1/samples/domains/helloworld_domain/defaultserver, where ORACLE_CEP_HOME refers to the main Oracle CEP installation directory, such as /oracle_cep.

2. Edit the startwlevs.sh script to add the -Djava.net.preferIPv4Stack=true property to the java command line:

   "$JAVA_HOME/bin/java" $JVM_ARGS $DEBUG_ARGS -Djava.net.preferIPv4Stack=true  -Dwlevs.home="$USER_INSTALL_DIR" -Dbea.home="$BEA_HOME" -jar "${USER_INSTALL_DIR}/bin/wlevs.jar" $ARGS
   

3. Save and close the startwlevs.sh script.

4. Start the affected Oracle CEP server using the modified startwlevs.sh script.

24.2.10 Multi-Byte Characters are not Supported by loadgen

The loadgen utility supports only ASCII characters. It does not support multi-byte character sets.

24.2.11 Oracle CEP Server Will not Start if the Keystore and Private Key Passwords are Different

The passwords for the keystore and the alias for the SSL private key must be the same. If they are not, the Oracle CEP Server will not start.

For more information, see Section 24.3.12, "Oracle CEP Server Will not Start if the Keystore and Private Key Passwords are Different".

24.3.1 Installer File Names are Incorrect

The installer file names in the following sections are incorrect:

• "Installing Oracle CEP in Graphical Mode" in the Oracle CEP Getting Started

• "Installing Oracle CEP in Console Mode" in the Oracle CEP Getting Started

Table 24-3 lists the correct installer file names for each supported platform.

24.3.2 Reference to Additional Oracle CEP Samples Missing

"Overview of the Samples Provided in the Distribution Kit" in the Oracle CEP Getting Started should include the following information:

"Additional Oracle CEP sample code can be found at http://www.oracle.com/technology/sample_code/products/event-driven-architecture".

24.3.3 Colt Aggregate Function Example Output Incomplete

"Functions: Colt Aggregate" in the Oracle CEP CQL Language Reference contains incomplete example output for the following functions:

• AUTOCORRELATION

• DURBINWATSON

• HARMONICMEAN

• SAMPLESTANDARDDEVIATION

• TRIMMEDMEAN

24.3.4 User-Defined Single Row Function Does not Implement an Interface

The introductory paragraph in "How to Implement a User-Defined Single-Row Function" in the Oracle CEP CQL Language Reference reads:

"You implement a user-defined single-row function by implementing a Java class that implements the SingleElementFunction interface."

It should read:

"You implement a user-defined single-row function by implementing a Java class that provides a public constructor and a public method that is invoked to execute the function."

Step 2 of this procedure should read:

"Compile the user-defined function Java implementation class and register the class in your Oracle CEP application assembly file as Example 24-2 shows.

<wlevs:processor id="testProcessor">
    <wlevs:listener ref="providerCache"/>
    <wlevs:listener ref="outputCache"/>
    <wlevs:cache-source ref="testCache"/>
    <wlevs:function function-name="mymod" exec-method="execute" />
        <bean class="com.bea.wlevs.example.function.MyMod"/>
    </wlevs:function>
</wlevs:processor>

Specify the method that is invoked to execute the function using the wlevs:function element exec-method attribute. This method must be public and must be uniquely identifiable by its name (that is, the method cannot have been overridden).

For more information, see "wlevs:function" in the Oracle CEP IDE Developer's Guide for Eclipse."

24.3.5 JMS Adapter Properties Topic Missing

"Viewing and Changing the Configuration of a Stage" in the Oracle CEP Visualizer User's Guide is missing a topic on the JMS Adapter stage.

24.3.6 Configuring Severity for an Individual Module Procedure Missing Some Component Name Constants

"Configuring Severity for an Individual Module" in the Oracle CEP Administrator's Guide is missing some of the valid logging component name constants you can use. Table 24-4 provides the complete list.

<May 26, 2009 5:53:49 PM PDT> <Info> <EventTrace> <BEA-000000> <Application [helloworld], Stage [helloworldOutputChannel] received insert event>

<May 26, 2009 6:02:34 PM PDT> <Debug> <EventTrace> <BEA-000000> <Application [helloworld], Stage [helloworldOutputChannel] received insert event [HelloWorldEvent: HelloWorld - the current time is: 6:02:34 PM]>

24.3.7 Variable Duration Non-Event Detection is not Supported in This Release

"DURATION Clause" in the Oracle CEP CQL Language Reference describes using the DURATION clause for non-event detection. This section should distinguish between the following:

• For fixed duration non-event detection, you use the DURATION clause with a constant value and time unit or just a constant value.

• For variable duration non-event detection, you use the DURATION clause with an arbitrary arithmetic expression.

• For recurring non-event detection, you use the DURATION clause with a MULTIPLES OF clause.

Fixed duration non-event pattern detection is supported in 11g Release 1 (11.1.1). When you create a query for fixed duration non-event detection, you may use the DURATION clause with constant value and time unit, such as DURATION 5 SECONDS, or just a constant value such as DURATION 5.

Variable duration non-event pattern detection is not supported in 11g Release 1 (11.1.1). That is, you may not use the DURATION clause with an arbitrary arithmetic expression, such as DURATION c1+4.

Recurring non-event pattern detection is supported in 11g Release 1 (11.1.1) but only for the fixed duration case. That is, you may use a DURATION clause with a MULTIPLES OF clause but only for a duration that is a constant value.

The title of "Non-Event Detection" in the Oracle CEP CQL Language Reference should read "Fixed Duration Non-Event Detection".

24.3.8 DURATION Clause Examples Incorrect

The following examples in "DURATION Clause" in the Oracle CEP CQL Language Reference are incorrect and should read as shown below:

• Example 15-8 "MATCH_RECOGNIZE with DURATION Clause Stream Output"

  Timestamp   Tuple Kind  Tuple11000:      +           10,8825000:      +           10,38000:      +           10,1850000:      +           10,1962000:      +           10, 
  

• Example 15-11 "MATCH_RECOGNIZE with DURATION MULTIPLES OF Clause Stream Output"

  Timestamp   Tuple Kind  Tuple 11000:     +           10,88,637 25000:     +           10,, 38000:     +           10,18,18 48000:     +           10,19,37 58000:     +           10,19,37 68000:     +           10,20,57 82000:     +           10,, 92000:     +           10,,102000:     +           10,,112000:     +           10,, 
  

The paragraph after Example 15-8:

"Now the further tuples that arrive until 11000 all match B. Since the duration is 10 we output a match at time 1000+10000=11000. Note that the last of the tuples that match B is chosen (the one with the value 15) since the greedy quantifier * in the pattern regular expression AB* suggests as many B matches as possible should be matched. Since the sequence of tuples from 1000 through 11000 match the pattern AB* and nothing else a match is reported at 11000."

Should read:

"Since the duration is 10 we output a match as soon as input at time 1000+10000=11000 is received (the one with the value 12). Since the sequence of tuples from 1000 through 9000 match the pattern AB* and nothing else a match is reported as soon as input at time 11000 is received."

24.3.9 Outer Join Operator "(+)" is Deprecated

"Comparison Conditions" in the Oracle CEP CQL Language Reference refers to the deprecated outer join operator (+). Oracle recommends that you not use the (+) join operator. Use the LEFT OUTER JOIN and RIGHT OUTER JOIN syntax described below.

"Outer Joins" in the Oracle CEP CQL Language Reference refers to the deprecated outer join operator (+). This section should read as follows:

An outer join extends the result of a simple join. An outer join returns all rows that satisfy the join condition and also returns some or all of those rows from one table for which no rows from the other satisfy the join condition.

Example 24-3 shows how to create a query q5 that uses a left outer join between streams S0, with schema (c1 integer, c2 float), and S1, with schema (c1 integer, c2 float).

<query id="q5"><![CDATA[ 
    SELECT a.c1+b.c1 
    FROM S0[range 5] AS a LEFT OUTER JOIN S1[range 3] AS b 
    ON b.c2 = a.c2
    WHERE b.c2 > 3
]]></query>

Use the ON clause to specify a join condition. Doing so lets you specify join conditions separate from any search or filter conditions in the WHERE clause.

You can perform the following types of outer join:

• Section 24.3.9.1, "Left Outer Join"

• Section 24.3.9.2, "Right Outer Join"

<query id="q5"><![CDATA[ 
    SELECT a.c1+b.c1 
    FROM S0[range 5] AS a LEFT OUTER JOIN S1[range 3] AS b 
    ON b.c2 = a.c2
    WHERE b.c2 > 3
]]></query>

<query id="q5"><![CDATA[ 
    SELECT a.c1+b.c1 
    FROM S0[range 5] AS a RIGHT OUTER JOIN S1[range 3] AS b 
    ON b.c2 = a.c2
    WHERE b.c2 > 3
]]></query>

24.3.10 Patch Client Section is Missing

"Installing Oracle CEP" in the Oracle CEP Getting Started is missing information on using the patch client.

For more information, see Oracle Smart Update Installing Patches and Maintenance Packs.

24.3.11 Avoiding a Single Point of Failure for Oracle CEP Visualizer Access

The procedures in "Creating an Oracle CEP Multi-Server Domain Using Oracle CEP Native Clustering" in the Oracle CEP Administrator's Guide and "Creating an Oracle CEP Multi-Server Domain Using Oracle Coherence" in the Oracle CEP Administrator's Guide contain steps that read:

"To improve performance, consider disabling Oracle CEP Visualizer access on all but one server in the domain."

Should read:

"To avoid a single point of failure, consider enabling Oracle CEP Visualizer on a small subset of n machines in the domain. Note that enabling Oracle CEP Visualizer on a given Oracle CEP Server may impact the performance of the server depending on the Oracle CEP Visualizer workload."

24.3.12 Oracle CEP Server Will not Start if the Keystore and Private Key Passwords are Different

"Creating an Oracle CEP Standalone-Server Domain Using the Configuration Wizard in Graphical Mode" in the Oracle CEP Administrator's Guide step 11 reads:

"Enter and confirm the password for the Oracle CEP domain identity keystore.

By default, the password for the certificate private key will be the same as the identity keystore; if you want it to be different, uncheck Use Keystore Password and enter the private key password."

Should read:

"Enter and confirm the password for the Oracle CEP domain identity keystore.

By default, the password for the certificate private key will be the same as the identity keystore; the Oracle CEP Server will not start if otherwise. Do not uncheck Use Keystore Password and do not enter a private key password."

"Configuring SSL to Secure Network Traffic" in the Oracle CEP Administrator's Guide is missing the following procedure:

24.3.13 Colt Functions DurbinWatson and sampleStandardDeviation are not Supported

Oracle CEP does not support Colt functions DurbinWatson and sampleStandardDeviation.

The following sections should be omitted:

• "DURBINWATSON" in the Oracle CEP CQL Language Reference

• "SAMPLESTANDARDDEVIATION" in the Oracle CEP CQL Language Reference

24.3.14 Oracle CQL Built-In Function Names Must be Lower Case

In the following sections, all built-in function names should be lower case:

• "Functions: Single-Row" in the Oracle CEP CQL Language Reference

• "Functions: Aggregate" in the Oracle CEP CQL Language Reference

• "Functions: Colt Single-Row" in the Oracle CEP CQL Language Reference

• "Functions: Colt Aggregate" in the Oracle CEP CQL Language Reference

• "Functions: java.lang.Math" in the Oracle CEP CQL Language Reference

24.3.15 Adapter Example Source Uses Deprecated API

In "Event Sources and Event Sinks" in the Oracle CEP IDE Developer's Guide for Eclipse, should use the API StreamSink and RelationSink and StreamSource and RelationSource instead of the deprecated EventSource and EventSink API.

Similarily, the following areas of the Oracle CEP IDE Developer's Guide for Eclipse should also use the API StreamSink and RelationSink and StreamSource and RelationSource instead of the deprecated EventSource and EventSink API:

• Section 24.3.15.1, "Implementing an Adapter or Event Bean as an Event Source"

• Section 24.3.15.2, "Implementing an Adapter or Event Bean as an Event Sink"

• Section 24.3.15.3, "com.bea.wlevs.management.Activate"

• Section 24.3.15.4, "com.bea.wlevs.management.Prepare"

• Section 24.3.15.5, "com.bea.wlevs.management.Rollback"

package com.bea.wlevs.adapter.example.helloworld;

import java.text.DateFormat;
import java.util.Date;

import com.bea.wlevs.ede.api.RunnableBean;
import com.bea.wlevs.ede.api.StreamSender;
import com.bea.wlevs.ede.api.StreamSource;
import com.bea.wlevs.event.example.helloworld.HelloWorldEvent;

public class HelloWorldAdapter implements RunnableBean, StreamSource {

    private static final int SLEEP_MILLIS = 300;

    private DateFormat dateFormat;
    private String message;
    private boolean suspended;

    private StreamSender eventSender;

    public HelloWorldAdapter() {
        super();
        dateFormat = DateFormat.getTimeInstance();
    }

    public void run() {
        suspended = false;
        while (!isSuspended()) { // Generate messages forever...

            generateHelloMessage();

            try {
                synchronized (this) {
                    wait(SLEEP_MILLIS);
                }
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }
    }

    public void setMessage(String message) {
        this.message = message;
    }

    private void generateHelloMessage() {
        String message = this.message + dateFormat.format(new Date());
        HelloWorldEvent event = new HelloWorldEvent();
        event.setMessage(message);
    
        eventSender.sendInsertEvent(event);
    }

    public void setEventSender(StreamSender sender) {
        eventSender = sender;
    }

    public synchronized void suspend() {
        suspended = true;
    }
 
    private synchronized boolean isSuspended() {
        return suspended;
    }
}

package com.bea.wlevs.example.helloworld;
 
import com.bea.wlevs.ede.api.StreamSink;
import com.bea.wlevs.event.example.helloworld.HelloWorldEvent;
 
public class HelloWorldBean implements StreamSink {
 
    public void onInsertEvent(Object event) {
        if (event instanceof HelloWorldEvent) {
            HelloWorldEvent helloWorldEvent = (HelloWorldEvent) event;
            System.out.println("Message: " + helloWorldEvent.getMessage());
        }   
    }
 
}

package com.bea.wlevs.adapter.example.helloworld;
...
import com.bea.wlevs.configuration.Activate;
import com.bea.wlevs.ede.api.RunnableBean;
import com.bea.wlevs.ede.api.StreamSender;
import com.bea.wlevs.ede.api.StreamSource;
import com.bea.wlevs.event.example.helloworld.HelloWorldEvent;

public class HelloWorldAdapter implements RunnableBean, StreamSource {
...
    @Activate
    public void activateAdapter(HelloWorldAdapterConfig adapterConfig) {
        this.message = adapterConfig.getMessage();
    }
...
}

package com.bea.wlevs.adapter.example.helloworld;
...
import com.bea.wlevs.configuration.Prepare;
import com.bea.wlevs.ede.api.RunnableBean;
import com.bea.wlevs.ede.api.StreamSender;
import com.bea.wlevs.ede.api.StreamSource;
import com.bea.wlevs.event.example.helloworld.HelloWorldEvent;

public class HelloWorldAdapter implements RunnableBean, StreamSource {
...
    @Prepare
    public void checkConfiguration(HelloWorldAdapterConfig adapterConfig) {
        if (adapterConfig.getMessage() == null
                || adapterConfig.getMessage().length() == 0) {
            throw new RuntimeException("invalid message: " + message);
        }
    }
...
}

package com.bea.wlevs.adapter.example.helloworld;
...
import com.bea.wlevs.configuration.Rollback;
import com.bea.wlevs.ede.api.RunnableBean;
import com.bea.wlevs.ede.api.StreamSender;
import com.bea.wlevs.ede.api.StreamSource;
import com.bea.wlevs.event.example.helloworld.HelloWorldEvent;

public class HelloWorldAdapter implements RunnableBean, StreamSource {
...
    @Rollback
    public void rejectConfigurationChange(HelloWorldAdapterConfig adapterConfig) {
    }