3. Using Spring AMQP

In this chapter, we will explore the interfaces and classes that are the essential components for developing applications with Spring AMQP.

3.1 AMQP Abstractions

Spring AMQP consists of a handful of modules, each represented by a JAR in the distribution. These modules are: spring-amqp, spring-rabbit and spring-erlang. The 'spring-amqp' module contains the org.springframework.amqp.core package. Within that package, you will find the classes that represent the core AMQP "model". Our intention is to provide generic abstractions that do not rely on any particular AMQP broker implementation or client library. End user code will be more portable across vendor implementations as it can be developed against the abstraction layer only. These abstractions are then used implemented by broker-specific modules, such as 'spring-rabbit'. For the 1.0 release there is only a RabbitMQ implementation however the abstractions have been validated in .NET using Apache Qpid in addition to RabbitMQ. Since AMQP operates at the protocol level in principle the RabbitMQ client can be used with any broker that supports the same protocol version, but we do not test any other brokers at present.

The overview here assumes that you are already familiar with the basics of the AMQP specification already. If you are not, then have a look at the resources listed in Part IV, “Other Resources”

3.1.1. Message

The 0-8 and 0-9-1 AMQP specifications do not define a Message class or interface. Instead, when performing an operation such as ' basicPublish ', the content is passed as a byte-array argument and additional properties are passed in as separate arguments. Spring AMQP defines a Message class as part of a more general AMQP domain model representation. The purpose of the Message class is to simply encapsulate the body and properties within a single instance so that the API can in turn be simpler. The Message class definition is quite straightforward.

public class Message {

    private final MessageProperties messageProperties;

    private final byte[] body;

    public Message(byte[] body, MessageProperties messageProperties) {
        this.body = body;
        this.messageProperties = messageProperties;
    }

    public byte[] getBody() {
        return this.body;
    }

    public MessageProperties getMessageProperties() {
        return this.messageProperties;
    }
}

The MessageProperties interface defines several common properties such as 'messageId', 'timestamp', 'contentType', and several more. Those properties can also be extended with user-defined 'headers' by calling the setHeader(String key, Object value) method.

3.1.2. Exchange

The Exchange interface represents an AMQP Exchange, which is what a Message Producer sends to. Each Exchange within a virtual host of a broker will have a unique name as well as a few other properties:

public interface Exchange {

    String getName();

    String getExchangeType();

    boolean isDurable();

    boolean isAutoDelete();

    Map<String, Object> getArguments();

}

As you can see, an Exchange also has a 'type' represented by constants defined in ExchangeTypes. The basic types are: Direct, Topic, Fanout, and Headers. In the core package you will find implementations of the Exchange interface for each of those types. The behavior varies across these Exchange types in terms of how they handle bindings to Queues. For example, a Direct exchange allows for a Queue to be bound by a fixed routing key (often the Queue's name). A Topic exchange supports bindings with routing patterns that may include the '*' and '#' wildcards for 'exactly-one' and 'zero-or-more', respectively. The Fanout exchange publishes to all Queues that are bound to it without taking any routing key into consideration. For much more information about these and the other Exchange types, check out Part IV, “Other Resources”.

[Note]Note

The AMQP specification also requires that any broker provide a "default" Direct Exchange that has no name. All Queues that are declared will be bound to that default Exchange with their names as routing keys. You will learn more about the default Exchange's usage within Spring AMQP in Section 3.3, “AmqpTemplate”.

3.1.3. Queue

The Queue class represents the component from which a Message Consumer receives Messages. Like the various Exchange classes, our implementation is intended to be an abstract representation of this core AMQP type.

public class Queue  {

    private final String name;

    private volatile boolean durable;

    private volatile boolean exclusive;

    private volatile boolean autoDelete;

    private volatile Map<String, Object> arguments;

    /**
     * The queue is durable, non-exclusive and non auto-delete.
     *
     * @param name the name of the queue.
     */
    public Queue(String name) {
        this(name, true, false, false);
    }

    // Getters and Setters omitted for brevity

Notice that the constructor takes the Queue name. Depending on the implementation, the admin template may provide methods for generating a uniquely named Queue. Such Queues can be useful as a "reply-to" address or other temporary situations. For that reason, the 'exclusive' and 'autoDelete' properties of an auto-generated Queue would both be set to 'true'.

[Note]Note
See the section on queues in Section 3.8, “Configuring the broker” for information about declaring queues using namespace support, including queue arguments.

3.1.4. Binding

Given that a producer sends to an Exchange and a consumer receives from a Queue, the bindings that connect Queues to Exchanges are critical for connecting those producers and consumers via messaging. In Spring AMQP, we define a Binding class to represent those connections. Let's review the basic options for binding Queues to Exchanges.

You can bind a Queue to a DirectExchange with a fixed routing key.

new Binding(someQueue, someDirectExchange, "foo.bar")

You can bind a Queue to a TopicExchange with a routing pattern.

new Binding(someQueue, someTopicExchange, "foo.*")

You can bind a Queue to a FanoutExchange with no routing key.

new Binding(someQueue, someFanoutExchange)

We also provide a BindingBuilder to facilitate a "fluent API" style.

Binding b = BindingBuilder.bind(someQueue).to(someTopicExchange).with("foo.*");
[Note]Note

The BindingBuilder class is shown above for clarity, but this style works well when using a static import for the 'bind()' method.

By itself, an instance of the Binding class is just holding the data about a connection. In other words, it is not an "active" component. However, as you will see later in Section 3.8, “Configuring the broker”, Binding instances can be used by the AmqpAdmin class to actually trigger the binding actions on the broker. Also, as you will see in that same section, the Binding instances can be defined using Spring's @Bean-style within @Configuration classes. There is also a convenient base class which further simplifies that approach for generating AMQP-related bean definitions and recognizes the Queues, Exchanges, and Bindings so that they will all be declared on the AMQP broker upon application startup.

The AmqpTemplate is also defined within the core package. As one of the main components involved in actual AMQP messaging, it is discussed in detail in its own section (see Section 3.3, “AmqpTemplate”).

3.2 Connection and Resource Management

Whereas the AMQP model we described in the previous section is generic and applicable to all implementations, when we get into the management of resources, the details are specific to the broker implementation. Therefore, in this section, we will be focusing on code that exists only within our "spring-rabbit" module since at this point, RabbitMQ is the only supported implementation.

The central component for managing a connection to the RabbitMQ broker is the ConnectionFactory interface. The responsibility of a ConnectionFactory implementation is to provide an instance of org.springframework.amqp.rabbit.connection.Connection which is a wrapper for com.rabbitmq.client.Connection. The only concrete implementation we provide is CachingConnectionFactory which establishes a single connection proxy that can be shared by the application. Sharing of the connection is possible since the "unit of work" for messaging with AMQP is actually a "channel" (in some ways, this is similar to the relationship between a Connection and a Session in JMS). As you can imagine, the connection instance provides a createChannel method. The CachingConnectionFactory implementation supports caching of those channels, and it maintains separate caches for channels based on whether they are transactional or not. When creating an instance of CachingConnectionFactory, the 'hostname' can be provided via the constructor. The 'username' and 'password' properties should be provided as well. If you would like to configure the size of the channel cache (the default is 1), you could call the setChannelCacheSize() method here as well.

CachingConnectionFactory connectionFactory = new CachingConnectionFactory("somehost");
connectionFactory.setUsername("guest");
connectionFactory.setPassword("guest");

Connection connection = connectionFactory.createConnection();

When using XML, the configuration might look like this:

<bean id="connectionFactory"
      class="org.springframework.amqp.rabbit.connection.CachingConnectionFactory">
    <constructor-arg value="somehost"/>
    <property name="username" value="guest"/>
    <property name="password" value="guest"/>
</bean>
[Note]Note
There is also a SingleConnectionFactory implementation which is only available in the unit test code of the framework. It is simpler than CachingConnectionFactory since it does not cache channels, but it is not intended for practical usage outside of simple tests due to its lack of performance and resilience. If you find a need to implement your own ConnectionFactory for some reason, the AbstractConnectionFactory base class may provide a nice starting point.

A ConnectionFactory can be created quickly and conveniently using the rabbit namespace:

<rabbit:connection-factory id="connectionFactory"/>

In most cases this will be preferable since the framework can choose the best defaults for you. The created instance will be a CachingConnectionFactory. Keep in mind that the default cache size for channels is 1. If you want more channels to be cached set a larger value via the 'channelCacheSize' property. In XML it would look like this:

<bean id="connectionFactory"
      class="org.springframework.amqp.rabbit.connection.CachingConnectionFactory">
    <constructor-arg value="somehost"/>
    <property name="username" value="guest"/>
    <property name="password" value="guest"/>
    <property name="channelCacheSize" value="25"/>
</bean>

And with the namespace you can just add the 'channel-cache-size' attribute:

<rabbit:connection-factory
    id="connectionFactory" channel-cache-size="25"/>

Host and port attributes can be provided using the namespace

<rabbit:connection-factory
    id="connectionFactory" host="somehost" port="5672"/>

Alternatively, if running in a clustered environment, use the addresses attribute.

<rabbit:connection-factory
    id="connectionFactory" addresses="host1:5672,host2:5672"/>

3.2.1 Configuring the Underlying Client Connection Factory

The CachingConnectionFactory uses an instance of the Rabbit client ConnectionFactory; a number of configuration properties are passed through (host, port, userName, password, requestedHeartBeat for example) when setting the equivalent property on the CachingConnectionFactory . To set other properties (clientProperties for example), define an instance of the rabbit factory and provide a reference to it using the appropriate constructor of the CachingConnectionFactory. When using the namespace as described above, provide a reference to the configured factory in the connection-factory attribute.

<rabbit:connection-factory
      id="connectionFactory" connection-factory="rabbitConnectionFactory"/>

3.2.2 Publisher Confirms and Returns

Confirmed and returned messages are supported by setting the CachingConnectionFactory's publisherConfirms and publisherReturns properties to 'true' respectively.

When these options are set, Channels created by the factory are wrapped in an PublisherCallbackChannel which is used to facilitate the callbacks. When such a channel is obtained, the client can register a PublisherCallbackChannel.Listener with the Channel. The PublisherCallbackChannel implementation contains logic to route a confirm/return to the appropriate listener.

These features are explained further in the following sections.

[Tip]Tip
For some more background information, please see the following blog post by the RabbitMQ team titled Introducing Publisher Confirms.

3.3 AmqpTemplate

As with many other high-level abstractions provided by the Spring Framework and related projects, Spring AMQP provides a "template" that plays a central role. The interface that defines the main operations is called AmqpTemplate. Those operations cover the general behavior for sending and receiving Messages. In other words, they are not unique to any implementation, hence the "AMQP" in the name. On the other hand, there are implementations of that interface that are tied to implementations of the AMQP protocol. Unlike JMS, which is an interface-level API itself, AMQP is a wire-level protocol. The implementations of that protocol provide their own client libraries, so each implementation of the template interface will depend on a particular client library. Currently, there is only a single implementation: RabbitTemplate. In the examples that follow, you will often see usage of an "AmqpTemplate", but when you look at the configuration examples, or any code excerpts where the template is instantiated and/or setters are invoked, you will see the implementation type (e.g. "RabbitTemplate").

As mentioned above, the AmqpTemplate interface defines all of the basic operations for sending and receiving Messages. We will explore Message sending and reception, respectively, in the two sections that follow.

3.3.1 Publisher Confirms and Returns

The RabbitTemplate implementation of AmqpTemplate supports Publisher Confirms and Returns.

For returned messages, the template's mandatory property must be set to 'true', and it requires a CachingConnectionFactory that has its publisherReturns property set to true (see Section 3.2.2, “Publisher Confirms and Returns”). Returns are sent to to the client by it registering a RabbitTemplate.ReturnCallback by calling setReturnCallback(ReturnCallback callback). The callback must implement this method:

void returnedMessage(Message message, int replyCode, String replyText,
          String exchange, String routingKey);

Only one ReturnCallback is supported by each RabbitTemplate.

For Publisher Confirms (aka Publisher Acknowledgements), the template requires a CachingConnectionFactory that has its publisherConfirms property set to true. Confirms are sent to to the client by it registering a RabbitTemplate.ConfirmCallback by calling setConfirmCallback(ConfirmCallback callback). The callback must implement this method:

void confirm(CorrelationData correlationData, boolean ack);

The CorrelationData is an object supplied by the client when sending the original message. This is described further in the next section.

Only one ConfirmCallback is supported by a RabbitTemplate.

3.4 Sending messages

When sending a Message, one can use any of the following methods:

void send(Message message) throws AmqpException;

void send(String routingKey, Message message) throws AmqpException;

void send(String exchange, String routingKey, Message message) throws AmqpException;

We can begin our discussion with the last method listed above since it is actually the most explicit. It allows an AMQP Exchange name to be provided at runtime along with a routing key. The last parameter is the callback that is responsible for actual creating of the Message instance. An example of using this method to send a Message might look this this:

amqpTemplate.send("marketData.topic", "quotes.nasdaq.FOO",
    new Message("12.34".getBytes(), someProperties));

The "exchange" property can be set on the template itself if you plan to use that template instance to send to the same exchange most or all of the time. In such cases, the second method listed above may be used instead. The following example is functionally equivalent to the previous one:

amqpTemplate.setExchange("marketData.topic");
amqpTemplate.send("quotes.nasdaq.FOO", new Message("12.34".getBytes(), someProperties));

If both the "exchange" and "routingKey" properties are set on the template, then the method accepting only the Message may be used:

amqpTemplate.setExchange("marketData.topic");
amqpTemplate.setRoutingKey("quotes.nasdaq.FOO");
amqpTemplate.send(new Message("12.34".getBytes(), someProperties));

A better way of thinking about the exchange and routing key properties is that the explicit method parameters will always override the template's default values. In fact, even if you do not explicitly set those properties on the template, there are always default values in place. In both cases, the default is an empty String, but that is actually a sensible default. As far as the routing key is concerned, it's not always necessary in the first place (e.g. a Fanout Exchange). Furthermore, a Queue may be bound to an Exchange with an empty String. Those are both legitimate scenarios for reliance on the default empty String value for the routing key property of the template. As far as the Exchange name is concerned, the empty String is quite commonly used because the AMQP specification defines the "default Exchange" as having no name. Since all Queues are automatically bound to that default Exchange (which is a Direct Exchange) using their name as the binding value, that second method above can be used for simple point-to-point Messaging to any Queue through the default Exchange. Simply provide the queue name as the "routingKey" - either by providing the method parameter at runtime:

RabbitTemplate template = new RabbitTemplate(); // using default no-name Exchange
template.send("queue.helloWorld", new Message("Hello World".getBytes(), someProperties));

Or, if you prefer to create a template that will be used for publishing primarily or exclusively to a single Queue, the following is perfectly reasonable:

RabbitTemplate template = new RabbitTemplate(); // using default no-name Exchange
template.setRoutingKey("queue.helloWorld"); // but we'll always send to this Queue
template.send(new Message("Hello World".getBytes(), someProperties));

3.4.1 Publisher Confirms

With the RabbitTemplate implementation of AmqpTemplate, each of the send() methods has an overloaded version that takes an additional CorrelationData object. When publisher confirms are enabled, this object is returned in the callback described in Section 3.3, “AmqpTemplate”. This allows the sender to correlate a confirm (ack or nack) with the sent message.

3.4.2 Publisher Returns

When the template's mandatory property is 'true' returned messages are provided by the callback described in Section 3.3, “AmqpTemplate”.

3.5 Receiving messages

Message reception is always a bit more complicated than sending. The reason is that there are two ways to receive a Message. The simpler option is to poll for a single Message at a time with a synchronous, blocking method call. The more complicated yet more common approach is to register a listener that will receive Messages on-demand, asynchronously. We will look at an example of each approach in the next two sub-sections.

3.5.1 Synchronous Consumer

The AmqpTemplate itself can be used for synchronous Message reception. There are two 'receive' methods available. As with the Exchange on the sending side, there is a method that requires a queue property having been set directly on the template itself, and there is a method that accepts a queue parameter at runtime.

Message receive() throws AmqpException;

Message receive(String queueName) throws AmqpException;

Just like in the case of sending messages, the AmqpTemplate has some convenience methods for receiving POJOs instead of Message instances, and implementations will provide a way to customize the MessageConverter used to create the Object returned:

Object receiveAndConvert() throws AmqpException;

Object receiveAndConvert(String queueName) throws AmqpException;

3.5.2 Asynchronous Consumer

For asynchronous Message reception, a dedicated component (not the AmqpTemplate) is involved. That component is a container for a Message consuming callback. We will look at the container and its properties in just a moment, but first we should look at the callback since that is where your application code will be integrated with the messaging system. There are a few options for the callback. The simplest of these is to implement the MessageListener interface:

public interface MessageListener {
    void onMessage(Message message);
}

If your callback logic depends upon the AMQP Channel instance for any reason, you may instead use the ChannelAwareMessageListener. It looks similar but with an extra parameter:

public interface ChannelAwareMessageListener {
    void onMessage(Message message, Channel channel) throws Exception;
}

If you prefer to maintain a stricter separation between your application logic and the messaging API, you can rely upon an adapter implementation that is provided by the framework. This is often referred to as "Message-driven POJO" support. When using the adapter, you only need to provide a reference to the instance that the adapter itself should invoke.

MessageListener listener = new MessageListenerAdapter(somePojo);

Now that you've seen the various options for the Message-listening callback, we can turn our attention to the container. Basically, the container handles the "active" responsibilities so that the listener callback can remain passive. The container is an example of a "lifecycle" component. It provides methods for starting and stopping. When configuring the container, you are essentially bridging the gap between an AMQP Queue and the MessageListener instance. You must provide a reference to the ConnectionFactory and the queue name or Queue instance(s) from which that listener should consume Messages. Here is the most basic example using the default implementation, SimpleMessageListenerContainer :

SimpleMessageListenerContainer container = new SimpleMessageListenerContainer();
container.setConnectionFactory(rabbitConnectionFactory);
container.setQueueNames("some.queue");
container.setMessageListener(new MessageListenerAdapter(somePojo));

As an "active" component, it's most common to create the listener container with a bean definition so that it can simply run in the background. This can be done via XML:

<rabbit:listener-container connection-factory="rabbitConnectionFactory">
    <rabbit:listener queues="some.queue" ref="somePojo" method="handle"/>
</rabbit:listener-container>

Or, you may prefer to use the @Configuration style which will look very similar to the actual code snippet above:

@Configuration
public class ExampleAmqpConfiguration {

    @Bean
    public MessageListenerContainer messageListenerContainer() {
        SimpleMessageListenerContainer container = new SimpleMessageListenerContainer();
        container.setConnectionFactory(rabbitConnectionFactory());
        container.setQueueName("some.queue");
        container.setMessageListener(exampleListener());
        return container;
    }

    @Bean
    public ConnectionFactory rabbitConnectionFactory() {
        CachingConnectionFactory connectionFactory =
            new CachingConnectionFactory("localhost");
        connectionFactory.setUsername("guest");
        connectionFactory.setPassword("guest");
        return connectionFactory;
    }

    @Bean
    public MessageListener exampleListener() {
        return new MessageListener() {
            public void onMessage(Message message) {
                System.out.println("received: " + message);
            }
        };
    }
}

3.6 Message Converters

The AmqpTemplate also defines several methods for sending and receiving Messages that will delegate to a MessageConverter. The MessageConverter itself is quite straightforward. It provides a single method for each direction: one for converting to a Message and another for converting from a Message. Notice that when converting to a Message, you may also provide properties in addition to the object. The "object" parameter typically corresponds to the Message body.

public interface MessageConverter {

    Message toMessage(Object object, MessageProperties messageProperties)
            throws MessageConversionException;

    Object fromMessage(Message message) throws MessageConversionException;

}

The relevant Message-sending methods on the AmqpTemplate are listed below. They are simpler than the methods we discussed previously because they do not require the Message instance. Instead, the MessageConverter is responsible for "creating" each Message by converting the provided object to the byte array for the Message body and then adding any provided MessageProperties.

void convertAndSend(Object message) throws AmqpException;

void convertAndSend(String routingKey, Object message) throws AmqpException;

void convertAndSend(String exchange, String routingKey, Object message)
    throws AmqpException;

void convertAndSend(Object message, MessagePostProcessor messagePostProcessor)
    throws AmqpException;

void convertAndSend(String routingKey, Object message,
    MessagePostProcessor messagePostProcessor) throws AmqpException;

void convertAndSend(String exchange, String routingKey, Object message,
    MessagePostProcessor messagePostProcessor) throws AmqpException;

On the receiving side, there are only two methods: one that accepts the queue name and one that relies on the template's "queue" property having been set.

Object receiveAndConvert() throws AmqpException;

Object receiveAndConvert(String queueName) throws AmqpException;
[Note]Note
The MessageListenerAdapter mentioned in Section 3.5.2, “Asynchronous Consumer” also uses a MessageConverter.

3.6.1. SimpleMessageConverter

The default implementation of the MessageConverter strategy is called SimpleMessageConverter. This is the converter that will be used by an instance of RabbitTemplate if you do not explicitly configure an alternative. It handles text-based content, serialized Java objects, and simple byte arrays.

Converting From a Message

If the content type of the input Message begins with "text" (e.g. "text/plain"), it will also check for the content-encoding property to determine the charset to be used when converting the Message body byte array to a Java String. If no content-encoding property had been set on the input Message, it will use the "UTF-8" charset by default. If you need to override that default setting, you can configure an instance of SimpleMessageConverter, set its "defaultCharset" property and then inject that into a RabbitTemplate instance.

If the content-type property value of the input Message is set to "application/x-java-serialized-object", the SimpleMessageConverter will attempt to deserialize (rehydrate) the byte array into a Java object. While that might be useful for simple prototyping, it's generally not recommended to rely on Java serialization since it leads to tight coupling between the producer and consumer. Of course, it also rules out usage of non-Java systems on either side. With AMQP being a wire-level protocol, it would be unfortunate to lose much of that advantage with such restrictions. In the next two sections, we'll explore some alternatives for passing rich domain object content without relying on Java serialization.

For all other content-types, the SimpleMessageConverter will return the Message body content directly as a byte array.

Converting To a Message

When converting to a Message from an arbitrary Java Object, the SimpleMessageConverter likewise deals with byte arrays, Strings, and Serializable instances. It will convert each of these to bytes (in the case of byte arrays, there is nothing to convert), and it will set the content-type property accordingly. If the Object to be converted does not match one of those types, the Message body will be null.

3.6.2. JsonMessageConverter and Jackson2JsonMessageConverter

As mentioned in the previous section, relying on Java serialization is generally not recommended. One rather common alternative that is more flexible and portable across different languages and platforms is JSON (JavaScript Object Notation). Two implementations are available and can be configured on any RabbitTemplate instance to override its usage of the SimpleMessageConverter default. The JsonMessageConverter which uses the org.codehaus.jackson 1.x library and Jackson2JsonMessageConverter which uses the com.fasterxml.jackson 2.x library.

<bean class="org.springframework.amqp.rabbit.core.RabbitTemplate">
    <property name="connectionFactory" ref="rabbitConnectionFactory"/>
    <property name="messageConverter">
        <bean class="org.springframework.amqp.support.converter.JsonMessageConverter">
            <!-- if necessary, override the DefaultClassMapper -->
            <property name="classMapper" ref="customClassMapper"/>
        </bean>
    </property>
</bean>
<bean class="org.springframework.amqp.rabbit.core.RabbitTemplate">
    <property name="connectionFactory" ref="rabbitConnectionFactory"/>
    <property name="messageConverter">
        <bean class="org.springframework.amqp.support.converter.Jackson2JsonMessageConverter">
            <!-- if necessary, override the DefaultClassMapper -->
            <property name="classMapper" ref="customClassMapper"/>
        </bean>
    </property>
</bean>

As shown above, the JsonMessageConverter and Jackson2JsonMessageConverter uses a DefaultClassMapper by default. Type information is added to (and retrieved from) the MessageProperties. If an inbound message does not contain type information in the MessageProperties, but you know the expected type, you can configure a static type using the defaultType property

<bean id="jsonConverterWithDefaultType"
      class="o.s.amqp.support.converter.JsonMessageConverter">
    <property name="classMapper">
        <bean class="org.springframework.amqp.support.converter.DefaultClassMapper">
            <property name="defaultType" value="foo.PurchaseOrder"/>
        </bean>
    </property>
</bean>
<bean id="jsonConverterWithDefaultType"
      class="o.s.amqp.support.converter.Jackson2JsonMessageConverter">
    <property name="classMapper">
        <bean class="org.springframework.amqp.support.converter.DefaultClassMapper">
            <property name="defaultType" value="foo.PurchaseOrder"/>
        </bean>
    </property>
</bean>

3.6.3. MarshallingMessageConverter

Yet another option is the MarshallingMessageConverter. It delegates to the Spring OXM library's implementations of the Marshaller and Unmarshaller strategy interfaces. You can read more about that library here. In terms of configuration, it's most common to provide the constructor argument only since most implementations of Marshaller will also implement Unmarshaller.

<bean class="org.springframework.amqp.rabbit.core.RabbitTemplate">
    <property name="connectionFactory" ref="rabbitConnectionFactory"/>
    <property name="messageConverter">
        <bean class="org.springframework.amqp.support.converter.MarshallingMessageConverter">
            <constructor-arg ref="someImplemenationOfMarshallerAndUnmarshaller"/>
        </bean>
    </property>
</bean>

3.7 Request/Reply Messaging

The AmqpTemplate also provides a variety of sendAndReceive methods that accept the same argument options that you have seen above for the one-way send operations (exchange, routingKey, and Message). Those methods are quite useful for request/reply scenarios since they handle the configuration of the necessary "reply-to" property before sending and can listen for the reply message on an exclusive Queue that is created internally for that purpose.

Similar request/reply methods are also available where the MessageConverter is applied to both the request and reply. Those methods are named convertSendAndReceive. See the Javadoc of AmqpTemplate for more detail.

By default, a new temporary queue is used for each reply. However, a single reply queue can be configured on the template, which can be more efficient, and also allows you to set arguments on that queue. In this case, however, you must also provide a <reply-listener/> sub element. This element provides a listener container for the reply queue, with the template being the listener. All of the Section 3.11, “Message Listener Container Configuration” attributes allowed on a <listener-container/> are allowed on the element, except for connection-factory and message-converter, which are inherited from the template's configuration.

<rabbit:template id="amqpTemplate"
        connection-factory="connectionFactory" reply-queue="replies">
    <rabbit:reply-listener/>
</rabbit:template>

While the container and template share a connection factory, they do not share a channel and therefore requests and replies are not performed within the same transaction (if transactional).

3.7.1 Message Correlation With A Reply Queue

When using a fixed reply queue, it is necessary to provide correlation data so that replies can be correlated to requests. See RabbitMQ Remote Procedure Call (RPC). By default, the standard correlationId property will be used to hold the correlation data. However, if you wish to use a custom propertry to hold correlation data, you can set the correlation-key attribute on the <rabbit-template/>. Explicitly setting the attribute to correlationId is the same as omitting the attribute. Of course, the client and server must use the same header for correlation data.

[Note]Note
Spring AMQP version 1.1 used a custom property spring_reply_correlation for this data. If you wish to revert to this behavior with the current version, perhaps to maintain compatibility with another application using 1.1, you must set the attribute to spring_reply_correlation.

Reply Listener Container

When using a fixed reply queue, a SimpleListenerContainer is used to receive the replies; with the RabbitTemplate being the MessageListener. When defining a template with the <rabbit:template/> namespace element, as shown above, the parser defines the container and wires in the template as the listener.

[Note]Note
When the template does not use a fixed replyQueue, a listener container is not needed.

If you define your RabbitTemplate as a <bean/>, or using an @Configuration class to define it as an @Bean, or when creating the template programmatically, you will need to define and wire up the reply listener container yourself. If you fail to do this, the template will never receive the replies and will eventually time out and return null as the reply to a call to a sendAndReceive method.

[Important]Important
When wiring the reply listener and template yourself, it is important to ensure that the template's replyQueue and the container's queues (or queueNames) properties refer to the same queue. The template inserts the reply queue into the outbound message replyTo property.

The following are examples of how to manually wire up the beans.

<bean id="amqpTemplate" class="org.springframework.amqp.rabbit.core.RabbitTemplate">
    <constructor-arg ref="connectionFactory" />
    <property name="exchange" value="foo.exchange" />
    <property name="routingKey" value="foo" />
    <property name="replyQueue" ref="replyQ" />
    <property name="replyTimeout" value="600000" />
</bean>

<bean class="org.springframework.amqp.rabbit.listener.SimpleMessageListenerContainer">
    <constructor-arg ref="connectionFactory" />
    <property name="queues" ref="replyQ" />
    <property name="messageListener" ref="amqpTemplate" />
</bean>

<rabbit:queue id="replyQ" name="my.reply.queue" />
    @Bean
    public RabbitTemplate amqpTemplate() {
        RabbitTemplate rabbitTemplate = new RabbitTemplate(connectionFactory());
        rabbitTemplate.setMessageConverter(msgConv());
        rabbitTemplate.setReplyQueue(replyQueue());
        rabbitTemplate.setReplyTimeout(60000);
        return rabbitTemplate;
    }

    @Bean
    public SimpleMessageListenerContainer replyListenerContainer() {
        SimpleMessageListenerContainer container = new SimpleMessageListenerContainer();
        container.setConnectionFactory(connectionFactory());
        container.setQueues(replyQueue());
        container.setMessageListener(amqpTemplate());
        return container;
    }

    @Bean
    public Queue replyQueue() {
        return new Queue("my.reply.queue");
    }

3.7.2 Spring Remoting with AMQP

The Spring Framework has a general remoting capability, allowing Remote Procedure Calls (RPC) using various transports. Spring-AMQP supports a similar mechanism with a AmqpProxyFactoryBean on the client and a AmqpInvokerServiceExporter on the server. This provides RPC over AMQP. On the client side, a RabbitTemplate is used as described above; on the server side, the invoker (configured as a MessageListener) receives the message, invokes the configured service, and returns the reply using the inbound message's replyTo information.

The client factory bean can be injected into any bean (using its serviceInterface); the client can then invoke methods on the proxy, resulting in remote execution over AMQP.

[Note]Note

With the default MessageConverters, the method paramters and returned value must be instances of Serializable.

On the server side, the AmqpInvokerServiceExporter has both AmqpTemplate and MessageConverter properties. Currently, the template's MessageConverter is not used. If you need to supply a custom message converter, then you should provide it using the messageConverter property. On the client side, a custom message converter can be added to the AmqpTemplate which is provided to the AmqpProxyFactoryBean using its amqpTemplate property.

Sample client and server configurations are shown below.

<bean id="client"
	class="org.springframework.amqp.remoting.client.AmqpProxyFactoryBean">
	<property name="amqpTemplate" ref="template" />
	<property name="serviceInterface" value="foo.ServiceInterface" />
</bean>

<rabbit:connection-factory id="connectionFactory" />

<rabbit:template id="template" connection-factory="connectionFactory" reply-timeout="2000"
	routing-key="remoting.binding" exchange="remoting.exchange" />

<rabbit:admin connection-factory="connectionFactory" />

<rabbit:queue name="remoting.queue" />

<rabbit:direct-exchange name="remoting.exchange">
	<rabbit:bindings>
		<rabbit:binding queue="remoting.queue" key="remoting.binding" />
	</rabbit:bindings>
</rabbit:direct-exchange>
<bean id="listener"
	class="org.springframework.amqp.remoting.service.AmqpInvokerServiceExporter">
	<property name="serviceInterface" value="foo.ServiceInterface" />
	<property name="service" ref="service" />
	<property name="amqpTemplate" ref="template" />
</bean>

<bean id="service" class="foo.ServiceImpl" />

<rabbit:connection-factory id="connectionFactory" />

<rabbit:template id="template" connection-factory="connectionFactory" />

<rabbit:queue name="remoting.queue" />

<rabbit:listener-container connection-factory="connectionFactory">
	<rabbit:listener ref="listener" queue-names="remoting.queue" />
</rabbit:listener-container>
[Important]Important
The AmqpInvokerServiceExporter can only process properly formed messages, such as those sent from the AmqpProxyFactoryBean. If it receives a message that it cannot interpret, a serialized RuntimeException will be sent as a reply. If the message has no replyToAddress property, the message will be rejected and permanently lost if no Dead Letter Exchange has been configured.

3.8 Configuring the broker

The AMQP specification describes how the protocol can be used to configure Queues, Exchanges and Bindings on the broker. These operations which are portable from the 0.8 specification and higher are present in the AmqpAdmin interface in the org.springframework.amqp.core package. The RabbitMQ implementation of that class is RabbitAdmin located in the org.springframework.amqp.rabbit.core package.

The AmqpAdmin interface is based on using the Spring AMQP domain abstractions and is shown below:

public interface AmqpAdmin {

    // Exchange Operations

    void declareExchange(Exchange exchange);

    void deleteExchange(String exchangeName);

    // Queue Operations

    Queue declareQueue();

    void declareQueue(Queue queue);

    void deleteQueue(String queueName);

    void deleteQueue(String queueName, boolean unused, boolean empty);

    void purgeQueue(String queueName, boolean noWait);

    // Binding Operations

    void declareBinding(Binding binding);

}

The no-arg declareQueue() method defines a queue on the broker whose name is automatically generated. The additional properties of this auto-generated queue are exclusive=true, autoDelete=true, and durable=false.

[Note]Note

Removing a binding was not introduced until the 0.9 version of the AMQP spec.

The RabbitMQ implementation of this interface is RabbitAdmin which when configured using Spring XML would look like this:

<rabbit:connection-factory id="connectionFactory"/>

<rabbit:admin id="amqpAdmin" connection-factory="connectionFactory"/>

The RabbitAdmin implementation does automatic lazy declaration of Queues, Exchanges and Bindings declared in the same ApplicationContext. These components will be declared as son as a Connection is opened to the broker. There are some namespace features that make this very convenient, e.g. in the Stocks sample application we have:

<rabbit:queue id="tradeQueue"/>

<rabbit:queue id="marketDataQueue"/>

<fanout-exchange name="broadcast.responses"
                 xmlns="http://www.springframework.org/schema/rabbit">
    <bindings>
        <binding queue="tradeQueue"/>
    </bindings>
</fanout-exchange>

<topic-exchange name="app.stock.marketdata"
                xmlns="http://www.springframework.org/schema/rabbit">
    <bindings>
        <binding queue="marketDataQueue" pattern="${stocks.quote.pattern}"/>
    </bindings>
</topic-exchange>

In the example above we are using anonymous Queues (actually internally just Queues with names generated by the framework, not by the broker) and refer to them by ID. We can also declare Queues with explicit names, which also serve as identifiers for their bean definitions in the context. E.g.

<rabbit:queue name="stocks.trade.queue"/>
[Tip]Tip
You can provide both an id and a name attribute. This allows you to refer to the queue (for example in a binding) by an id that is independent of the queue name. It also allows standard Spring features such as property placeholders, and SpEL expressions for the queue name; these features are not available when using the name as the bean identifier.

Queues can be configured with additional arguments, for example, 'x-message-ttl' or 'x-ha-policy'. Using the namespace support, they are provided in the form of a Map of argument name/argument value pairs, using the <rabbit:queue-arguments> element.

<rabbit:queue name="withArguments">
    <rabbit:queue-arguments>
        <entry key="x-ha-policy" value="all"/>
    </rabbit:queue-arguments>
</rabbit:queue>

By default, the arguments are assumed to be strings. For arguments of other types, the type needs to be provided.

<rabbit:queue name="withArguments">
    <rabbit:queue-arguments value-type="java.lang.Long">
        <entry key="x-message-ttl" value="100"/>
    </rabbit:queue-arguments>
</rabbit:queue>

When providing arguments of mixed types, the type is provided for each entry element:

<rabbit:queue name="withArguments">
    <rabbit:queue-arguments>
        <entry key="x-message-ttl">
            <value type="java.lang.Long">100</value>
        </entry>
        <entry key="x-ha-policy" value="all"/>
    </rabbit:queue-arguments>
</rabbit:queue>

With Spring Framework 3.2 and later, this can be declared a little more succinctly:

<rabbit:queue name="withArguments">
    <rabbit:queue-arguments>
        <entry key="x-message-ttl" value="100" value-type="java.lang.Long"/>
        <entry key="x-ha-policy" value="all"/>
    </rabbit:queue-arguments>
</rabbit:queue>
[Important]Important

The RabbitMQ broker will not allow declaration of a queue with mismatched arguments. For example, if a queue already exists with no time to live argument, and you attempt to declare it with, say, key="x-message-ttl" value="100", an exception will be thrown.

By default, the RabbitAdmin will immediately stop processing all declarations when any exception occurs; this could cause downstream issues - such as a listener container failing to initialize because another queue (defined after the one in error) is not declared.

This behavior can be modified by setting the ignore-declaration-failures attribute to true on the RabbitAdmin. This option instructs the RabbitAdmin to log the exception, and continue declaring other elements.

To see how to use Java to configure the AMQP infrastructure, look at the Stock sample application, where there is the @Configuration class AbstractStockRabbitConfiguration which in turn has RabbitClientConfiguration and RabbitServerConfiguration subclasses. The code for AbstractStockRabbitConfiguration is shown below

@Configuration
public abstract class AbstractStockAppRabbitConfiguration {

    @Bean
    public ConnectionFactory connectionFactory() {
        CachingConnectionFactory connectionFactory =
            new CachingConnectionFactory("localhost");
        connectionFactory.setUsername("guest");
        connectionFactory.setPassword("guest");
        return connectionFactory;
    }

    @Bean
    public RabbitTemplate rabbitTemplate() {
        RabbitTemplate template = new RabbitTemplate(connectionFactory());
        template.setMessageConverter(jsonMessageConverter());
        configureRabbitTemplate(template);
        return template;
    }

    @Bean
    public MessageConverter jsonMessageConverter() {
        return new JsonMessageConverter();
    }

    @Bean
    public TopicExchange marketDataExchange() {
        return new TopicExchange("app.stock.marketdata");
    }

    // additional code omitted for brevity

}

In the Stock application, the server is configured using the following @Configuration class:

@Configuration
public class RabbitServerConfiguration extends AbstractStockAppRabbitConfiguration  {

    @Bean
    public Queue stockRequestQueue() {
        return new Queue("app.stock.request");
    }
}

This is the end of the whole inheritance chain of @Configuration classes. The end result is the the TopicExchange and Queue will be declared to the broker upon application startup. There is no binding of the TopicExchange to a queue in the server configuration, as that is done in the client application. The stock request queue however is automatically bound to the AMQP default exchange - this behavior is defined by the specification.

The client @Configuration class is a little more interesting and is shown below.

@Configuration
public class RabbitClientConfiguration extends AbstractStockAppRabbitConfiguration {

    @Value("${stocks.quote.pattern}")
    private String marketDataRoutingKey;

    @Bean
    public Queue marketDataQueue() {
        return amqpAdmin().declareQueue();
    }

    /**
     * Binds to the market data exchange. Interested in any stock quotes
     * that match its routing key.
     */
    @Bean
    public Binding marketDataBinding() {
        return BindingBuilder.bind(
                marketDataQueue()).to(marketDataExchange()).with(marketDataRoutingKey);
    }

    // additional code omitted for brevity

}

The client is declaring another queue via the declareQueue() method on the AmqpAdmin, and it binds that queue to the market data exchange with a routing pattern that is externalized in a properties file.

3.8.1 Conditional Declaration

By default, all queues, exchanges, and bindings are declared by all RabbitAdmin instances (that have auto-startup="true") in the application context.

[Note]Note
Starting with the 1.2 release, it is possible to conditionally declare these elements. This is particularly useful when an application connects to multiple brokers and needs to specify with which broker(s) a particular element should be declared.

The classes representing these elements implement Declarable which has two methods: shouldDeclare() and getDeclaringAdmins(). The RabbitAdmin uses these methods to determine whether a particular instance should actually process the declarations on its Connection.

The properties are available as attributes in the namespace, as shown in the following examples.

<rabbit:admin id="admin1" connection-factory="CF1" />

<rabbit:admin id="admin2" connection-factory="CF2" />

<rabbit:queue id="declaredByBothAdminsImplicitly" />

<rabbit:queue id="declaredByBothAdmins" declared-by="admin1, admin2" />

<rabbit:queue id="declaredByAdmin1Only" declared-by="admin1" />

<rabbit:queue id="notDeclaredByAny" auto-declare="false" />

<rabbit:direct-exchange name="direct" declared-by="admin1, admin2">
	<rabbit:bindings>
		<rabbit:binding key="foo" queue="bar"/>
	</rabbit:bindings>
</rabbit:direct-exchange>
[Note]Note
The auto-declare attribute is true by default and if the declared-by is not supplied (or is empty) then all RabbitAdmins will declare the object (as long as the admin's auto-startup attribute is true; the default).

Similarly, you can use Java-based @Configuration to achieve the same effect. In this example, the components will be declared by admin1 but not admin2:

@Bean
public RabbitAdmin admin() {
	RabbitAdmin rabbitAdmin = new RabbitAdmin(cf1());
	rabbitAdmin.afterPropertiesSet();
	return rabbitAdmin;
}

@Bean
public RabbitAdmin admin2() {
	RabbitAdmin rabbitAdmin = new RabbitAdmin(cf2());
	rabbitAdmin.afterPropertiesSet();
	return rabbitAdmin;
}

@Bean
public Queue queue() {
	Queue queue = new Queue("foo");
	queue.setAdminsThatShouldDeclare(admin());
	return queue;
}

@Bean
public Exchange exchange() {
	DirectExchange exchange = new DirectExchange("bar");
	exchange.setAdminsThatShouldDeclare(admin());
	return exchange;
}

@Bean
public Binding binding() {
	Binding binding = new Binding("foo", DestinationType.QUEUE, exchange().getName(), "foo", null);
	binding.setAdminsThatShouldDeclare(admin());
	return binding;
}

3.9 Exception Handling

Many operations with the RabbitMQ Java client can throw checked Exceptions. For example, there are a lot of cases where IOExceptions may be thrown. The RabbitTemplate, SimpleMessageListenerContainer, and other Spring AMQP components will catch those Exceptions and convert into one of the Exceptions within our runtime hierarchy. Those are defined in the 'org.springframework.amqp' package, and AmqpException is the base of the hierarchy.

If you are using a SimpleMessageListenerContainer you will also be able to inject a Spring ErrorHandler instance that can be used to react to an exception in the listener. The ErrorHandler can influence whether or not the errant message is requeued by throwing a AmqpRejectAndDontRequeueException. See ??? for more information about this exception.

3.10 Transactions

The Spring Rabbit framework has support for automatic transaction management in the synchronous and asynchronous use cases with a number of different semantics that can be selected declaratively, as is familiar to existing users of Spring transactions. This makes many if not most common messaging patterns very easy to implement.

There are two ways to signal the desired transaction semantics to the framework. In both the RabbitTemplate and SimpleMessageListenerContainer there is a flag channelTransacted which, if true, tells the framework to use a transactional channel and to end all operations (send or receive) with a commit or rollback depending on the outcome, with an exception signaling a rollback. Another signal is to provide an external transaction with one of Spring's PlatformTransactionManager implementations as a context for the ongoing operation. If there is already a transaction in progress when the framework is sending or receiving a message, and the channelTransacted flag is true, then the commit or rollback of the messaging transaction will be deferred until the end of the current transaction. If the channelTransacted flag is false, then no transaction semantics apply to the messaging operation (it is auto-acked).

The channelTransacted flag is a configuration time setting: it is declared and processed once when the AMQP components are created, usually at application startup. The external transaction is more dynamic in principle because the system responds to the current Thread state at runtime, but in practice is often also a configuration setting, when the transactions are layered onto an application declaratively.

For synchronous use cases with RabbitTemplate the external transaction is provided by the caller, either declaratively or imperatively according to taste (the usual Spring transaction model). An example of a declarative approach (usually preferred because it is non-invasive), where the template has been configured with channelTransacted=true:

@Transactional
public void doSomething() {
    String incoming = rabbitTemplate.receiveAndConvert();
    // do some more database processing...
    String outgoing = processInDatabaseAndExtractReply(incoming);
    rabbitTemplate.convertAndSend(outgoing);
}

A String payload is received, converted and sent as a message body inside a method marked as @Transactional, so if the database processing fails with an exception, the incoming message will be returned to the broker, and the outgoing message will not be sent. This applies to any operations with the RabbitTemplate inside a chain of transactional methods (unless the Channel is directly manipulated to commit the transaction early for instance).

For asynchronous use cases with SimpleMessageListenerContainer if an external transaction is needed it has to be requested by the container when it sets up the listener. To signal that an external transaction is required the user provides an implementation of PlatformTransactionManager to the container when it is configured. For example:

@Configuration
public class ExampleExternalTransactionAmqpConfiguration {

    @Bean
    public MessageListenerContainer messageListenerContainer() {
        SimpleMessageListenerContainer container = new SimpleMessageListenerContainer();
        container.setConnectionFactory(rabbitConnectionFactory());
        container.setTransactionManager(transactionManager());
        container.setChannelTransacted(true);
        container.setQueueName("some.queue");
        container.setMessageListener(exampleListener());
        return container;
    }

}

In the example above, the transaction manager is added as a dependency injected from another bean definition (not shown), and the channelTransacted flag is also set to true. The effect is that if the listener fails with an exception the transaction will be rolled back, and the message will also be returned to the broker. Significantly, if the transaction fails to commit (e.g. a database constraint error, or connectivity problem), then the AMQP transaction will also be rolled back, and the message will be returned to the broker. This is sometimes known as a Best Efforts 1 Phase Commit, and is a very powerful pattern for reliable messaging. If the channelTransacted flag was set to false in the example above, which is the default, then the external transaction would still be provided for the listener, but all messaging operations would be auto-acked, so the effect is to commit the messaging operations even on a rollback of the business operation.

3.10.1 A note on Rollback of Received Messages

AMQP transactions only apply to messages and acks sent to the broker, so when there is a rollback of a Spring transaction and a message has been received, what Spring AMQP has to do is not just rollback the transaction, but also manually reject the message (sort of a nack, but that's not what the specification calls it). Such messages (and any that are unacked when a channel is closed or aborts) go to the back of the queue on a Rabbit broker, and this behaviour is not what some users expect, especially if they come from a JMS background, so it's good to be aware of it. The re-queuing order is not mandated by the AMQP specification, but it makes the broker much more efficient, and also means that if it is under load there is a natural back off before the message can be consumed again.

3.10.2 Using the RabbitTransactionManager

The RabbitTransactionManager is an alternative to executing Rabbit operations within, and synchronized with, external transactions. This Transaction Manager is an implementation of the PlatformTransactionManager interface and should be used with a single Rabbit ConnectionFactory.

[Important]Important
This strategy is not able to provide XA transactions, for example in order to share transactions between messaging and database access.

Application code is required to retrieve the transactional Rabbit resources via ConnectionFactoryUtils.getTransactionalResourceHolder(ConnectionFactory, boolean) instead of a standard Connection.createChannel() call with subsequent Channel creation. When using Spring's RabbitTemplate, it will autodetect a thread-bound Channel and automatically participate in it.

With Java Configuration you can setup a new RabbitTransactionManager using:

@Bean
public RabbitTransactionManager rabbitTransactionManager() {
    return new RabbitTransactionManager(connectionFactory);
}

If you prefer using XML configuration, declare the following bean in your XML Application Context file:

<bean id="rabbitTxManager"
      class="org.springframework.amqp.rabbit.transaction.RabbitTransactionManager">
    <property name="connectionFactory" ref="connectionFactory"/>
</bean>

3.11 Message Listener Container Configuration

There are quite a few options for configuring a SimpleMessageListenerContainer related to transactions and quality of service, and some of them interact with each other.

When configuring with the XML namespace, the convention is to use hyphenated attributes rather than camel case; for example, for property 'connectionFactory', the XML equivalent is 'connection-factory'.

Table 3.1. Configuration options for a message listener container

PropertyDescription
channelTransactedBoolean flag to signal that all messages should be acknowledged in a transaction (either manually or automatically)
acknowledgeModeNONE = no acks will be sent (incompatible with channelTransacted=true). RabbitMQ calls this "autoack" because the broker assumes all messages are acked without any action from the consumer. MANUAL = the listener must acknowledge all messages by calling Channel.basicAck(). AUTO = the container will acknowledge the message automatically, unless the MessageListener throws an exception. Note that acknowledgeMode is complementary to channelTransacted - if the channel is transacted then the broker requires a commit notification in addition to the ack. This is the default mode. See also txSize.
transactionManagerExternal transaction manager for the operation of the listener. Also complementary to channelTransacted - if the Channel is transacted then its transaction will be synchronized with the external transaction.
prefetchCountThe number of messages to accept from the broker in one socket frame. The higher this is the faster the messages can be delivered, but the higher the risk of non-sequential processing. Ignored if the acknowledgeMode is NONE. This will be increased, if necessary, to match the txSize.
shutdownTimeoutWhen a container shuts down (e.g. if its enclosing ApplicationContext is closed) it waits for in-flight messages to be processed up to this limit. Defaults to 10 seconds. After the limit is reached, if the channel is not transacted messages will be discarded.
txSizeWhen used with acknowledgeMode AUTO, the container will attempt to process up to this number of messages before sending an ack (waiting for each one up to the receive timeout setting). This is also when a transactional channel is committed. If the prefetchCount is less than the txSize, it will be increased to match the txSize.
receiveTimeoutThe maximum time to wait for each message. If acknowledgeMode=NONE this has very little effect - the container just spins round and asks for another message. It has the biggest effect for a transactional Channel with txSize > 1, since it can cause messages already consumed not to be acknowledged until the timeout expires.
autoStartupFlag to indicate that the container should start when the ApplicationContext does (as part of the SmartLifecycle callbacks which happen after all beans are initialized). Defaults to true, but set it to false if your broker might not be available on startup, and then call start() later manually when you know the broker is ready.
phaseWhen autoStartup is true, the lifecycle phase within which this container should start and stop. The lower the value the earlier this container will start and the later it will stop. The default is Integer.MAX_VALUE meaning the container will start as late as possible and stop as soon as possible.
adviceChainAn array of AOP Advice to apply to the listener execution. This can be used to apply additional cross cutting concerns such as automatic retry in the event of broker death. Note that simple re-connection after an AMQP error is handled by the CachingConnectionFactory, as long as the broker is still alive.
taskExecutorA reference to a Spring TaskExecutor (or standard JDK 1.5+ Executor) for executing listener invokers. Default is a SimpleAsyncTaskExecutor, using internally managed threads.
errorHandlerA reference to an ErrorHandler strategy for handling any uncaught Exceptions that may occur during the execution of the MessageListener.
concurrencyThe number of concurrent consumers to start for each listener.
connectionFactoryA reference to the connectionFactory; when configuring using the XML namespace, the default referenced bean name is "rabbitConnectionFactory".
messageConverterA reference to the MessageConverter strategy for converting AMQP Messages to listener method arguments for any referenced 'listener' that is a POJO. Default is a SimpleMessageConverter.
requeueRejectedDetermines whether messages that are rejected because the listener threw an exception should be requeued or not. Default 'true'.


3.12 Resilience: Recovering from Errors and Broker Failures

Some of the key (and most popular) high-level features that Spring AMQP provides are to do with recovery and automatic re-connection in the event of a protocol error or broker failure. We have seen all the relevant components already in this guide, but it should help to bring them all together here and call out the features and recovery scenarios individually.

The primary reconnection features are enabled by the CachingConnectionFactory itself. It is also often beneficial to use the RabbitAdmin auto-declaration features. In addition, if you care about guaranteed delivery, you probably also need to use the channelTransacted flag in RabbitTemplate and SimpleMessageListenerContainer and also the AcknowledgeMode.AUTO (or manual if you do the acks yourself) in the SimpleMessageListenerContainer.

3.12.1 Automatic Declaration of Exchanges, Queues and Bindings

The RabbitAdmin component can declare exchanges, queues and bindings on startup. It does this lazily, through a ConnectionListener, so if the broker is not present on startup it doesn't matter. The first time a Connection is used (e.g. by sending a message) the listener will fire and the admin features will be applied. A further benefit of doing the auto declarations in a listener is that if the connection is dropped for any reason (e.g. broker death, network glitch, etc.) they will be applied again the next time they are needed.

3.12.2 Failures in Synchronous Operations and Options for Retry

If you lose your connection to the broker in a synchronous sequence using RabbitTemplate (for instance), then Spring AMQP will throw an AmqpException (usually but not always AmqpIOException). We don't try to hide the fact that there was a problem, so you have to be able to catch and respond to the exception. The easiest thing to do if you suspect that the connection was lost, and it wasn't your fault, is to simply try the operation again. You can do this manually, or you could look at using Spring Retry to handle the retry (imperatively or declaratively).

Spring Retry provides a couple of AOP interceptors and a great deal of flexibility to specify the parameters of the retry (number of attempts, exception types, backoff algorithm etc.). Spring AMQP also provides some convenience factory beans for creating Spring Retry interceptors in a convenient form for AMQP use cases, with strongly typed callback interfaces for you to implement custom recovery logic. See the Javadocs and properties of StatefulRetryOperationsInterceptor and StatelessRetryOperationsInterceptor for more detail. Stateless retry is appropriate if there is no transaction or if a transaction is started inside the retry callback. Note that stateless retry is simpler to configure and analyse than stateful retry, but it is not usually appropriate if there is an ongoing transaction which must be rolled back or definitely is going to roll back. A dropped connection in the middle of a transaction should have the same effect as a rollback, so for reconnection where the transaction is started higher up the stack, stateful retry is usually the best choice.

3.12.3 Message Listeners and the Asynchronous Case

If a MessageListener fails because of a business exception, the exception is handled by the message listener container and then it goes back to listening for another message. If the failure is caused by a dropped connection (not a business exception), then the consumer that is collecting messages for the listener has to be cancelled and restarted. The SimpleMessageListenerContainer handles this seamlessly, and it leaves a log to say that the listener is being restarted. In fact it loops endlessly trying to restart the consumer, and only if the consumer is very badly behaved indeed will it give up. One side effect is that if the broker is down when the container starts, it will just keep trying until a connection can be established.

Business exception handling, as opposed to protocol errors and dropped connections, might need more thought and some custom configuration, especially if transactions and/or container acks are in use. Prior to 2.8.x, RabbitMQ had no definition of dead letter behaviour, so by default a message that is rejected or rolled back because of a business exception can be redelivered ad infinitum. To put a limit in the client on the number of re-deliveries, one choice is a StatefulRetryOperationsInterceptor in the advice chain of the listener. The interceptor can have a recovery callback that implements a custom dead letter action: whatever is appropriate for your particular environment.

Another alternative is to set the container's rejectRequeued property to false. This causes all failed messages to be discarded. When using RabbitMQ 2.8.x or higher, this also facilitates delivering the message to a Dead Letter Exchange.

Or, you can throw a AmqpRejectAndDontRequeueException; this prevents message requeuing, regardless of the setting of the defaultRequeueRejected property.

Often, a combination of both techniques will be used. Use a StatefulRetryOperationsInterceptor in the advice chain, where it's MessageRecover throws an AmqpRejectAndDontRequeueException. The MessageRecover is called when all retries have been exhausted. The default MessageRecoverer simply consumes the errant message and emits a WARN message. In which case, the message is ACK'd and won't be sent to the Dead Letter Exchange, if any.

3.12.4 Exception Classification for Retry

Spring Retry has a great deal of flexibility for determining which exceptions can invoke retry. The default configuration will retry for all exceptions. Given that user exceptions will be wrapped in a ListenerExecutionFailedException we need to ensure that the classification examines the exception causes. The default classifier just looks at the top level exception.

Since Spring Retry 1.0.3, the BinaryExceptionClassifier has a property traverseCauses (default false). When true it will traverse exception causes until it finds a match or there is no cause.

To use this classifier for retry, use a SimpleRetryPolicy created with the constructor that takes the max attempts, the Map of Exceptions and the boolean (traverseCauses), and inject this policy into the RetryTemplate.