Last-Value queues are special queues which discard any messages when a newer message with the same value for a well-defined Last-Value property is put in the queue. In other words, a Last-Value queue only retains the last value.

A typical example for Last-Value queue is for stock prices, where you are only interested by the latest value for a particular stock.

Messages sent to an Last-Value queue without the specified property will be delivered as normal and will never be "replaced".

1. Configuration

Last Value Key Configuration

Last-Value queues can be statically configured in broker.xml via the last-value-key

<address name="foo.bar">
   <multicast>
      <queue name="orders1" last-value-key="reuters_code" />
   </multicast>
</address>

Specified on creating a queue by using the CORE api specifying the parameter lastValue to true.

Or on auto-create when using the JMS Client by using address parameters when creating the destination used by the consumer.

Queue queue = session.createQueue("my.destination.name?last-value-key=reuters_code");
Topic topic = session.createTopic("my.destination.name?last-value-key=reuters_code");

Address wildcards can be used to configure Last-Value queues for a set of addresses (see here).

<address-setting match="lastValueQueue">
   <default-last-value-key>reuters_code</default-last-value-key>
</address-setting>

By default, default-last-value-key is null.

Legacy Last Value Configuration

Last-Value queues can also just be configured via the last-value boolean property, doing so it will default the last-value-key to _AMQ_LVQ_NAME.

<address name="foo.bar">
   <multicast>
      <queue name="orders1" last-value="true" />
   </multicast>
</address>

Specified on creating a queue by using the CORE api specifying the parameter lastValue to true.

Or on auto-create when using the JMS Client by using address parameters when creating the destination used by the consumer.

Queue queue = session.createQueue("my.destination.name?last-value=true");
Topic topic = session.createTopic("my.destination.name?last-value=true");

Also the default for all queues under and address can be defaulted using the address-setting configuration:

<address-setting match="lastValueQueue">
   <default-last-value-queue>true</default-last-value-queue>
</address-setting>

By default, default-last-value-queue is false.

Note that address-setting last-value-queue config is deprecated, please use default-last-value-queue instead.

2. Last-Value Property

The property name used to identify the last value is configurable at the queue level mentioned above.

If using the legacy setting to configure an LVQ then the default property "_AMQ_LVQ_NAME" is used (or the constant Message.HDR_LAST_VALUE_NAME from the Core API).

For example, using the sample configuration

<address name="foo.bar">
   <multicast>
      <queue name="orders1" last-value-key="reuters_code" />
   </multicast>
</address>

if two messages with the same value for the Last-Value property are sent to a Last-Value queue, only the latest message will be kept in the queue:

// send 1st message with Last-Value property `reuters_code` set to `VOD`
TextMessage message = session.createTextMessage("1st message with Last-Value property set");
message.setStringProperty("reuters_code", "VOD");
producer.send(message);

// send 2nd message with Last-Value property `reuters_code` set to `VOD`
message = session.createTextMessage("2nd message with Last-Value property set");
message.setStringProperty("reuters_code", "VOD");
producer.send(message);

...

// only the 2nd message will be received: it is the latest with
// the Last-Value property set
TextMessage messageReceived = (TextMessage)messageConsumer.receive(5000);
System.out.format("Received message: %s\n", messageReceived.getText());

3. Forcing all consumers to be non-destructive

It’s common to combine last-value queues with non-destructive semantics.

4. Clustering

The fundamental ideas behind last-value queues and clustering are at odds with each other.

Clustering was designed as a way to increase message throughput through horizontal scaling. The messages in a clustered queue can be spread across all nodes in the cluster. This allows clients to be distributed across the cluster to leverage the computing resources all the nodes rather than being bottlenecked on a single node.

However, if you wanted to use a last-value queue in a cluster then in order to enforce last-value semantics all messages would be required to go to a queue on a single node. This would effectively nullify the benefits of clustering. Also, the arrival of messages on and and redistribution of those messages from nodes other than the node where the last-value semantics would be enforced would almost certainly impact which message is considered "last."

For these reasons last-value queues are not supported in a traditional cluster. However, it would be possible to use a connection router in front of a cluster (or even a set of non-clustered brokers) to ensure all clients which need to use the same last-value queue are directed to the same node. See the connection router for more details on configuration, etc.

5. Example

See the last-value queue example which shows how last value queues are configured and used with JMS.