org.apache.qpid.client.failover

Class FailoverRetrySupport<T,E extends Exception>

java.lang.Object

org.apache.qpid.client.failover.FailoverRetrySupport<T,E>

All Implemented Interfaces:

FailoverSupport<T,E>

public class FailoverRetrySupport<T,E extends Exception>
extends Object
implements FailoverSupport<T,E>

FailoverRetrySupport is a continuation that wraps another continuation, delaying its execution until it is notified that a blocking condition has been met, and executing the continuation within a mutex. If the continuation fails, due to the original condition being broken, whilst the continuation is waiting for a reponse to a synchronous request, FailoverRetrySupport automatcally rechecks the condition and re-acquires the mutex and re-runs the continution. This automatic retrying is continued until the continuation succeeds, or throws an exception (different to FailoverException, which is used to signal the failure of the original condition).

The blocking condition used is that the connection is not currently failing over, and the mutex used is the connection failover mutex, which guards against the fail-over process being run during fail-over vulnerable methods. These are used like a lock and condition variable.

The wrapped operation may throw a FailoverException, this is an exception that can be raised by a BlockingMethodFrameListener, in response to it being notified that a fail-over wants to start whilst it was waiting. Methods that are vulnerable to fail-over are those that are synchronous, where a failure will prevent them from getting the reply they are waiting for and asynchronous methods that should not be attempted when a fail-over is in progress.

Wrapping a synchronous method in a FailoverRetrySupport will have the effect that the operation will not be started during fail-over, but be delayed until any current fail-over has completed. Should a fail-over process want to start whilst waiting for the synchronous reply, the FailoverRetrySupport will detect this and retry the operation until it succeeds. Synchronous methods are usually coordinated with a BlockingMethodFrameListener which is notified when a fail-over process wants to start and throws a FailoverException in response to this.

Wrapping an asynchronous method in a FailoverRetrySupport will have the effect that the operation will not be started during fail-over, but be delayed until any current fail-over has completed.

TODO InterruptedException not handled well.

Constructor Summary

Constructors

Constructor and Description
FailoverRetrySupport(FailoverProtectedOperation<T,E> operation, AMQConnection con) Creates an automatic retrying fail-over handler for the specified operation.

Method Summary

Methods

Modifier and Type	Method and Description
T	execute() Delays a continuation until the "not failing over" condition is met on the specified connection.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

FailoverRetrySupport

public FailoverRetrySupport(FailoverProtectedOperation<T,E> operation,
                    AMQConnection con)

Creates an automatic retrying fail-over handler for the specified operation.

Parameters:

operation - The fail-over protected operation to wrap in this handler.

Method Detail

execute

public T execute()
          throws E extends Exception

Delays a continuation until the "not failing over" condition is met on the specified connection. Repeats until the operation throws AMQException or succeeds without being interrupted by fail-over.

Specified by:

execute in interface FailoverSupport<T,E extends Exception>

Returns:

The result of executing the continuation.

Throws:

E - Any underlying exception is allowed to fall through.

E extends Exception