Custom Exceptions

For situations where there is an unrecoverable error, the Business Process Management solution uses a combination of exception handlers and custom exceptions.

Handling Exceptions

Rather than using Terminate shapes in called orchestrations you can use the pattern of throwing exceptions that are handled by the root orchestration. This allows the orchestration with the widest knowledge of context to make the decision about handling the exception. Having the subordinate orchestrations throw custom exceptions enables you to communicate more specific information to the root orchestration.

The business process management solution follows this pattern. There are four root, or uncalled, orchestrations in the solution: OrderBroker, OrderManager, CableOrder1, and CableOrder2. Three of the root orchestrations receive and handle custom exceptions: OrderManager, CableOrder1, and CableOrder2.

Note that some of the root orchestrations use the Terminate shape and that the Terminate shape appears just before the normal end point of the orchestration. The orchestration still uses the Terminate shape in case of an error so that the orchestration is recorded as terminated, rather than as completed but with an error message. This allows the solution to track failed instances easily in addition to recording the error.

For more information about the Terminate shape, see How to Configure the Terminate Shape. For more information about the ThrowException shape, see How to Configure the Throw Exception Shape.

The Custom Exceptions

Each of the following three custom exceptions follows the same pattern. Having distinct types allows the code to distinguish different exceptions.


Exception Thrown By (Orchestration)




Activate, CableOrder1


CableOrder1, CheckInterrupt, ErrorHandlerOrch

All of the custom exceptions are defined in the Utilities assembly. The custom exceptions are all .NET classes. All of the custom exception classes inherit from the .NET ApplicationException class which in turn inherits from the System.Exception class. Because there is a possibility that the exception may be dehydrated (serialized and stored in the database), the exceptions must implement a deserialization constructor. A deserialization constructor is a constructor that takes two arguments: a SerializationInfo object, and a StreamingContext object. The deserialization constructor will be used during rehydration of the exception. Because the ApplicationException class already implements a deserialization constructor, the constructor for the custom exception simply invokes the base (ApplicationException) deserialization constructor. For example, the InterruptException includes the following constructor:

// "info" is the object holding the serialized object data.
// "context" is the contextual information about the source
// or destination.
public InterruptException(SerializationInfo info,
                  StreamingContext context) : base(info, context) { }

For more detailed information about custom serialization, see Custom Serialization in the .NET Framework Developer's Guide.

Nested Exception Handlers and Compensation Blocks

In addition to serving as specialized exceptions, the custom exceptions in several places also serve as a sort flag. In the Change orchestration, there are two places that the Cancel operation must be rolled back.

You may want to read this section with the Change orchestration open in Visual Studio.

At the top of the orchestration, if the call to the Cancel orchestration fails, the CancelCompensation block executes and rolls back the Cancel transaction. If all goes well, the orchestration then calls the Activate orchestration.

If the Activate operation fails, the Cancel transaction must also be rolled back. However, the exception handler for the call to Activate knows nothing about the Cancel transaction. To handle this, there is an exception handler for the entire orchestration. This handler, because it contains the Cancel scope, knows about the Cancel transaction. The handler catches the exception thrown from the Activate scope (the ActivateException), rolls back the Cancel transaction, and then throws the exception again so that the orchestration that called the Change orchestration can perform any additional processing.

Notice that this design only compensates the Cancel if the Activate fails. If the Cancel fails and throws an exception, the Cancel never took place and should not be compensated.

For more information about compensation blocks and the Compensate shape, see How to Configure the Compensate Shape.

See Also

  © 2009 Microsoft Corporation. All rights reserved.