Very good link given below .
http://javaoraclesoa.blogspot.com/2012/05/exception-handling-in-soa-suite-10g-and.html
Exception handling in SOA Suite 10g and SOA Suite 11g
Introduction
Sometimes, the longer you think about how to solve a problem, the less complex the solution becomes. Error handling in SOA Suite 11g is one of those examples. It is tempting to implement an own mechanism for exception/error handling (for example http://javaoraclesoa.blogspot.com/2012/05/re-enqueueing-faulted-bpel-messages.html), although there already is an extensive fault management framework part of the SOA Suite. In this post I describe the method used in SOA Suite 10g to implement fault-policies using a custom Java class. I implement a similar exception handling mechanism in Oracle SOA Suite 11g.
Marcel Bellinga has provided most code in the below example.
Challenges to tackle
Some of the challenges involved when implementing exception handling;
- how do I make it easy for the people monitoring and maintaining the application to detect and recover from errors?
- how do I make sure no messages are lost?
- how do I make sure the order in which messages are offered to the application, does not change when exceptions occur?
- how do I prevent 'hammering' a system (continuously retrying faulted messages)
With these questions in mind, the following solution provides a good option.
A bit of background
Oracle BPEL 10g has the option to use fault-policies and fault-bindings (and use custom Java classes in the policies), which are put on the application server and referred to by a bpel process in the bpel.xml. See; http://docs.oracle.com/cd/E14101_01/doc.1013/e15342/bpelrn.htm#BABCHCED.
Oracle SOA Suite 11g has (in addition to the method described above) the option to deploy custom Java classes, fault-policies and fault-bindings as part of the composite to the application server. This mechanism makes it easier to use the fault management framework on a per-composite basis. See http://docs.oracle.com/cd/E12839_01/integration.1111/e10224/bp_faults.htm
Keep in mind, when using the fault management framework that the fault-policies get triggered before a catch branch as defined in a BPEL process. If you want the catch branch to be activated, the action to rethrow the fault, needs to be part of the policy.
Solution in short
The solution for handling faults while taking into account the above questions, will use the following method;
- in Oracle BPEL 10g, a custom Java class and a specific policy xml-file is deployed on the application server
- the bpel.xml file will refer to the policy defined in the specific policy XML file
- the custom Java class will first deactivate the activation agents of the process and then retire the process (avoiding the issue that messages are picked up while the process is already retired causing loss of messages)
- the faulted message is put in manual recovery mode so the error hospital can be used to recover (retry) the message after the problem is fixed
- if the problem is fixed, the process can be activated again
- the ORABPEL schema tables can be monitored for messages which can be recovered or to trigger someone something has gone wrong and a recovery action is required
In Oracle SOA Suite 11g the method is similar, however, the activation agents do not need to be deactivated explicitly, the API calls are a bit different (due to the SCA implementation) and the error handling is deployed as part of the composite (in this example, see http://mazanatti.info/index.php?/archives/75-SOA-Fault-Framework-Creating-and-using-a-Java-action-fault-policy.html for an example on how to deploy custom Java code centrally on the server).
Implementation
Implementation BPEL 10g exception handling
Custom Java action
Create a new Java project and include the orabpel.jar from your BPEL distribution in the root folder of your project. Update the project libraries to include the library.
Create a new Java class. I've used the following;
package testapi;
import com.oracle.bpel.client.BPELProcessMetaData;
import com.oracle.bpel.client.IBPELProcessConstants;
import com.oracle.bpel.client.IBPELProcessHandle;
import com.oracle.bpel.client.Locator;
import com.oracle.bpel.client.config.faultpolicy.IFaultRecoveryContext;
import com.oracle.bpel.client.config.faultpolicy.IFaultRecoveryJavaClass;
public class RetireProcess implements IFaultRecoveryJavaClass {
public RetireProcess() {
}
/** * This method is called by the BPEL Error Hospital framework when this
* * action is selected as retrySuccessAction (with the retry option) or
* * when this action is selected as successor in the human intervention
* * screen in the BPEL Console. *
* * @param iFaultRecoveryContext */
public void handleRetrySuccess(IFaultRecoveryContext iFaultRecoveryContext) {
System.out.println("RetireProcess RetrySucces start");
setLifeCycle(iFaultRecoveryContext,
IBPELProcessConstants.LIFECYCLE_ACTIVE);
System.out.println("RetireProcess RetrySucces einde");
}
/** * This method is called by the BPEL Error Hospital framework when this
* * class is configured as action in the fault handling policy *
* * @param iFaultRecoveryContext
* * @return String that can be used to influence choice for next action (not used in this case) */
public String handleBPELFault(IFaultRecoveryContext iFaultRecoveryContext) {
System.out.println("RetireProcess HandleFault start");
setLifeCycle(iFaultRecoveryContext,
IBPELProcessConstants.LIFECYCLE_RETIRED);
System.out.println("RetireProcess HandleFault Einde");
return null;
}
private void setLifeCycle(IFaultRecoveryContext iFaultRecoveryContext,
int status) {
IBPELProcessHandle procHandle = null;
Locator loc = null;
BPELProcessMetaData bpelProcessMetadata = null;
String processName;
String revision;
try {
processName = iFaultRecoveryContext.getProcessId().getProcessId();
revision = iFaultRecoveryContext.getProcessId().getRevisionTag();
/*
* get Locator Instance
*/
loc = iFaultRecoveryContext.getLocator();
/*
* Lookup Process. Revision optional.
*/
if (revision == null || revision.trim().equals("")) {
procHandle = loc.lookupProcess(processName);
} else {
procHandle = loc.lookupProcess(processName, revision);
}
if (procHandle == null) {
throw new Exception("Unable to find process: " + processName);
}
System.out.println("RetireProcess set lifecycle to retired");
/*
* Get Metadata of the process.
*/
bpelProcessMetadata = procHandle.getMetaData();
if (bpelProcessMetadata.getLifecycle() != status) {
/*
* Set Lifecycle to Retired.
* Use setState(IBPELProcessConstants.STATE_OFF) to change process state to off.
*/
bpelProcessMetadata.setLifecycle(status);
System.out.println("RetireProcess Lifecycle set to retired");
/*
* Stop activation agents
*/
if (status == IBPELProcessConstants.LIFECYCLE_RETIRED) {
procHandle.stopAllActivationAgents();
} else {
procHandle.startAllActivationAgents();
;
}
/*
* Finally update the process with the modified metadata.
*/
procHandle.updateMetaData(bpelProcessMetadata);
}
} catch (Exception e) {
e.printStackTrace();
}
}
}
Noteworthy here are the method to retire the process; obtain a locator, use the locator to get a processhandle, use the processhandle to get to the metadata, update the metadata. The processhandle can also be used to stop the activation agents. Compile the project using JDK 1.5.0.06.
Place this class in;
[ORACLE_HOME]/bpel/system/classes/
Fault policy and fault binding
Create a fault policy like for example
8
2
Place the fault policy in
[ORACLE_HOME}/bpel/domains/{domain}/config/fault-policies
Create a reference to the faultpolicy in the bpel.xml of the process like;
(below )
Noteworthy in this policy is the defaultAction. My custom Java class returns null. This triggers the defaultAction which is set to ora-human-intervention. This causes the invoke to be visible in the error hospital (Activities tab in the process manager). From the error hospital it is also possible to specify an on retry success method to be executed (by clicking the specific error).
Result
When an error occurs, the failed messages arrive (in order) in the error hospital (usually a small number before the process is retired). The process instances which have faulted, remain open. The process is retired. You can retry the activities to check whether the error is fixed. In the error is fixed, the process can be activated again resuming normal action. This way the order of messages is guaranteed, there is no useless hammering and retrying the action which has failed. The process can be activated when the problem is fixed avoiding a lot of manual re-offering of messages.
Implementation BPEL 11g Exception handling
The 11g implementation is very similar to the 10g implementation. Deployment does not require any server side configuration. You can download the example project here; http://dl.dropbox.com/u/6693935/blog/TestExceptionHandling.zip. If you encounter errors deploying the project, you should remove the MDS entry in .adf\META-INF\adf-config.xml causing the issue. The example project requires the setup as described in; http://javaoraclesoa.blogspot.com/2012/05/re-enqueueing-faulted-bpel-messages.html. Also mind that when importing the project, your MDS configuration might differ. Remove the entries not relevant for your configuration from the .adf/META-INF/adf-config.xml file.
Custom Java class
I've used the following Java class (created in SCA-INF/src). No additional project configuration (like including libraries) is required in 11g.
package ms.testapp.exceptionhandling;
import com.collaxa.cube.engine.fp.BPELFaultRecoveryContextImpl;
import java.util.logging.Logger;
import oracle.integration.platform.faultpolicy.IFaultRecoveryContext;
import oracle.integration.platform.faultpolicy.IFaultRecoveryJavaClass;
import oracle.soa.management.facade.Composite;
import oracle.soa.management.facade.Locator;
import oracle.soa.management.facade.LocatorFactory;
public class RetireProcess implements IFaultRecoveryJavaClass {
private final static Logger logger = Logger.getLogger(RetireProcess.class.getName());
public RetireProcess() {
super();
}
public void handleRetrySuccess(IFaultRecoveryContext iFaultRecoveryContext) {
}
public String handleFault(IFaultRecoveryContext iFaultRecoveryContext) {
System.out.println("handleFault started");
BPELFaultRecoveryContextImpl bpelCtx =
(BPELFaultRecoveryContextImpl)iFaultRecoveryContext;
try{
Locator loc = LocatorFactory.createLocator();
System.out.println("locator obtained");
Composite comp = loc.lookupComposite(bpelCtx.getProcessDN().getCompositeDN());
System.out.println("compisite found");
comp.retire();
//bpelCtx.addAuditTrailEntry("retired " + comp.getDN());
System.out.println("process retired");
logger.info("retired " + comp.getDN());
} catch (Exception e) {
System.out.println("fault in handler");
//bpelCtx.addAuditTrailEntry("Error in FaultHandler " + RetireProcess.class.getName());
logger.severe("Error in FaultHandler " + RetireProcess.class.getName());
e.printStackTrace();
}
return null;
}
}
Fault policy and fault binding
My fault-policy file is called fault-policies.xml (the composite.xml picks that file by default but a different file can be specified in the composite.xml file if required) and it looks like;
My fault-bindings.xml looks like;
These files are placed in the same folder as the composite.xml.
Result
The behavior in 11g is similar to the behavior described in 10g in both examples. One thing to notice is that the API works on composite level and I've not found a way to directly stop of start the activation agents. I did however not encounter the 10g error that the JCA adapter tried to start a retired process.
First the correct situation. Use the testscript to enqueue a message.
DECLARE
queue_options DBMS_AQ.ENQUEUE_OPTIONS_T;
message_properties DBMS_AQ.MESSAGE_PROPERTIES_T;
recipients DBMS_AQ.aq$_recipient_list_t;
message_id RAW(16);
message SYS.XMLType;
BEGIN
recipients(1) := sys.aq$_agent('EXCEPTIONTEST', NULL, NULL);
message_properties.recipient_list := recipients;
message := sys.XMLType.createXML('Name Piet
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment