Note
	Prior to version 4.2 metrics were only available when JMX was enabled. See Section 9.2, “JMX Support”.

To enable MessageSource, MessageChannel and MessageHandler metrics, add an <int:management/> bean to the application context, or annotate one of your @Configuration classes with @EnableIntegrationManagement. MessageSource s only maintain counts, MessageChannel s and MessageHandler s maintain duration statistics in addition to counts. See Section 9.1.2, “MessageChannel Metric Features” and Section 9.1.3, “MessageHandler Metric Features” below.

This causes the automatic registration of the IntegrationManagementConfigurer bean in the application context. Only one such bean can exist in the context and it must have the bean name integrationManagementConfigurer if registered manually via a <bean/> definition.

In addition to metrics, you can control debug logging in the main message flow. It has been found that in very high volume applications, even calls to isDebugEnabled() can be quite expensive with some logging subsystems. You can disable all such logging to avoid this overhead; exception logging (debug or otherwise) are not affected by this setting.

A number of options are available:

<int:management
    default-logging-enabled="false" 
    default-counts-enabled="false" 
    default-stats-enabled="false" 
    counts-enabled-patterns="foo, !baz, ba*" 
    stats-enabled-patterns="fiz, buz" 
    metrics-factory="myMetricsFactory" /> 

@Configuration
@EnableIntegration
@EnableIntegrationManagement(
    defaultLoggingEnabled = "false", 
    defaultCountsEnabled = "false", 
    defaultStatsEnabled = "false", 
    countsEnabled = { "foo", "${count.patterns}" }, 
    statsEnabled = { "qux", "!*" }, 
    MetricsFactory = "myMetricsFactory") 
public static class ContextConfiguration {
...
}

At runtime, counts and statistics can be obtained by calling IntegrationManagementConfigurer getChannelMetrics, getHandlerMetrics and getSourceMetrics, returning MessageChannelMetrics, MessageHandlerMetrics and MessageSourceMetrics respectively.

See the javadocs for complete information about these classes.

When JMX is enabled (see Section 9.2, “JMX Support”), these metrics are also exposed by the IntegrationMBeanExporter.

Message channels report metrics according to their concrete type. If you are looking at a DirectChannel, you will see statistics for the send operation. If it is a QueueChannel, you will also see statistics for the receive operation, as well as the count of messages that are currently buffered by this QueueChannel. In both cases there are some metrics that are simple counters (message count and error count), and some that are estimates of averages of interesting quantities. The algorithms used to calculate these estimates are described briefly in the section below.

The following table shows the statistics maintained for message handlers. Some metrics are simple counters (message count and error count), and one is an estimate of averages of send duration. The algorithms used to calculate these estimates are described briefly in the table below:

A feature of the time-based average estimates is that they decay with time if no new measurements arrive. To help interpret the behaviour over time, the time (in seconds) since the last measurement is also exposed as a metric.

There are two basic exponential models: decay per measurement (appropriate for duration and anything where the number of measurements is part of the metric), and decay per time unit (more suitable for rate measurements where the time in between measurements is part of the metric). Both models depend on the fact that

S(n) = sum(i=0,i=n) w(i) x(i) has a special form when w(i) = r^i, with r=constant:

S(n) = x(n) + r S(n-1) (so you only have to store S(n-1), not the whole series x(i), to generate a new metric estimate from the last measurement). The algorithms used in the duration metrics use r=exp(-1/M) with M=10. The net effect is that the estimate S(n) is more heavily weighted to recent measurements and is composed roughly of the last M measurements. So M is the "window" or lapse rate of the estimate In the case of the vanilla moving average, i is a counter over the number of measurements. In the case of the rate we interpret i as the elapsed time, or a combination of elapsed time and a counter (so the metric estimate contains contributions roughly from the last M measurements and the last T seconds).

A new strategy interface MetricsFactory has been introduced allowing you to provide custom channel metrics for your MessageChannel s and MessageHandler s. By default, a DefaultMetricsFactory provides default implementation of MessageChannelMetrics and MessageHandlerMetrics which are described in the next bullet. To override the default MetricsFactory configure it as described above, by providing a reference to your MetricsFactory bean instance. You can either customize the default implementations as described in the next bullet, or provide completely different implementations by extending AbstractMessageChannelMetrics and/or AbstractMessageHandlerMetrics.

In addition to the default metrics factory described above, the framework provides the AggregatingMetricsFactory. This factory creates AggregatingMessageChannelMetrics and AggregatingMessageHandlerMetrics. In very high volume scenarios, the cost of capturing statistics can be prohibitive (2 calls to the system time and storing the data for each message). The aggregating metrics aggregate the response time over a sample of messages. This can save significant CPU time.

	Caution
	The statistics will be skewed if messages arrive in bursts. These metrics are intended for use with high, constant-volume, message rates.

<bean id="aggregatingMetricsFactory"
            class="org.springframework.integration.support.management.AggregatingMetricsFactory">
    <constructor-arg value="1000" /> <!-- sample size -->
</bean>

The above configuration aggregates the duration over 1000 messages. Counts (send, error) are maintained per-message but the statistics are per 1000 messages.

See Section 9.1.4, “Time-Based Average Estimates” and the Javadocs for the ExponentialMovingAverage* classes for more information about these values.

By default, the DefaultMessageChannelMetrics and DefaultMessageHandlerMetrics use a window of 10 measurements, a rate period of 1 second (rate per second) and a decay lapse period of 1 minute.

If you wish to override these defaults, you can provide a custom MetricsFactory that returns appropriately configured metrics and provide a reference to it to the MBean exporter as described above.

Example:

public static class CustomMetrics implements MetricsFactory {

    @Override
    public AbstractMessageChannelMetrics createChannelMetrics(String name) {
        return new DefaultMessageChannelMetrics(name,
                new ExponentialMovingAverage(20, 1000000.),
                new ExponentialMovingAverageRate(2000, 120000, 30, true),
                new ExponentialMovingAverageRatio(130000, 40, true),
                new ExponentialMovingAverageRate(3000, 140000, 50, true));
    }

    @Override
    public AbstractMessageHandlerMetrics createHandlerMetrics(String name) {
        return new DefaultMessageHandlerMetrics(name, new ExponentialMovingAverage(20, 1000000.));
    }

}

The customizations described above are wholesale and will apply to all appropriate beans exported by the MBean exporter. This is the extent of customization available using XML configuration.

Individual beans can be provided with different implementations using java @Configuration or programmatically at runtime, after the application context has been refreshed, by invoking the configureMetrics methods on AbstractMessageChannel and AbstractMessageHandler.

Previously, the time-based metrics (see Section 9.1.4, “Time-Based Average Estimates”) were calculated in real time. The statistics are now calculated when retrieved instead. This resulted in a significant performance improvement, at the expense of a small amount of additional memory for each statistic. As discussed in the bullet above, the statistics can be disabled altogether, while retaining the MBean allowing the invocation of Lifecycle methods.

The Notification-listening Channel Adapter requires a JMX ObjectName for the MBean that publishes notifications to which this listener should be registered. A very simple configuration might look like this:

<int-jmx:notification-listening-channel-adapter id="adapter"
    channel="channel"
    object-name="example.domain:name=publisher"/>

	Tip
	The notification-listening-channel-adapter registers with an MBeanServer at startup, and the default bean name is mbeanServer which happens to be the same bean name generated when using Spring’s <context:mbean-server/> element. If you need to use a different name, be sure to include the_mbean-server_ attribute.

The adapter can also accept a reference to a NotificationFilter and a handback Object to provide some context that is passed back with each Notification. Both of those attributes are optional. Extending the above example to include those attributes as well as an explicit MBeanServer bean name would produce the following:

<int-jmx:notification-listening-channel-adapter id="adapter"
    channel="channel"
    mbean-server="someServer"
    object-name="example.domain:name=somePublisher"
    notification-filter="notificationFilter"
    handback="myHandback"/>

The Notification-listening Channel Adapter is event-driven and registered with the MBeanServer directly. It does not require any poller configuration.

Note

For this component only, the object-name attribute can contain an ObjectName pattern (e.g. "org.foo:type=Bar,name=*") and the adapter will receive notifications from all MBeans with ObjectNames that match the pattern. In addition, the object-name attribute can contain a SpEL reference to a <util:list/> of ObjectName patterns: <jmx:notification-listening-channel-adapter id="manyNotificationsAdapter" channel="manyNotificationsChannel" object-name="#{patterns}"/> <util:list id="patterns"> <value>org.foo:type=Foo,name=*</value> <value>org.foo:type=Bar,name=*</value> </util:list> The names of the located MBean(s) will be logged when DEBUG level logging is enabled.

The Notification-publishing Channel Adapter is relatively simple. It only requires a JMX ObjectName in its configuration as shown below.

<context:mbean-export/>

<int-jmx:notification-publishing-channel-adapter id="adapter"
    channel="channel"
    object-name="example.domain:name=publisher"/>

It does also require that an MBeanExporter be present in the context. That is why the <context:mbean-export/> element is shown above as well.

When Messages are sent to the channel for this adapter, the Notification is created from the Message content. If the payload is a String it will be passed as the message text for the Notification. Any other payload type will be passed as the userData of the Notification.

JMX Notifications also have a type, and it should be a dot-delimited String. There are two ways to provide the type. Precedence will always be given to a Message header value associated with the JmxHeaders.NOTIFICATION_TYPE key. On the other hand, you can rely on a fallback default-notification-type attribute provided in the configuration.

<context:mbean-export/>

<int-jmx:notification-publishing-channel-adapter id="adapter"
    channel="channel"
    object-name="example.domain:name=publisher"
    default-notification-type="some.default.type"/>

The Attribute Polling Channel Adapter is useful when you have a requirement, to periodically check on some value that is available through an MBean as a managed attribute. The poller can be configured in the same way as any other polling adapter in Spring Integration (or it’s possible to rely on the default poller). The object-name and attribute-name are required. An MBeanServer reference is also required, but it will automatically check for a bean named mbeanServer by default, just like the Notification-listening Channel Adapter described above.

<int-jmx:attribute-polling-channel-adapter id="adapter"
    channel="channel"
    object-name="example.domain:name=someService"
    attribute-name="InvocationCount">
        <int:poller max-messages-per-poll="1" fixed-rate="5000"/>
</int-jmx:attribute-polling-channel-adapter>

The Tree Polling Channel Adapter queries the JMX MBean tree and sends a message with a payload that is the graph of objects that matches the query. By default the MBeans are mapped to primitives and simple Objects like Map, List and arrays - permitting simple transformation, for example, to JSON. An MBeanServer reference is also required, but it will automatically check for a bean named mbeanServer by default, just like the Notification-listening Channel Adapter described above. A basic configuration would be:

<int-jmx:tree-polling-channel-adapter id="adapter"
    channel="channel"
    query-name="example.domain:type=*">
        <int:poller max-messages-per-poll="1" fixed-rate="5000"/>
</int-jmx:tree-polling-channel-adapter>

This will include all attributes on the MBeans selected. You can filter the attributes by providing an MBeanObjectConverter that has an appropriate filter configured. The converter can be provided as a reference to a bean definition using the converter attribute, or as an inner <bean/> definition. A DefaultMBeanObjectConverter is provided which can take a MBeanAttributeFilter in its constructor argument.

Two standard filters are provided; the NamedFieldsMBeanAttributeFilter allows you to specify a list of attributes to include and the NotNamedFieldsMBeanAttributeFilter allows you to specify a list of attributes to exclude. You can also implement your own filter

The operation-invoking-channel-adapter enables Message-driven invocation of any managed operation exposed by an MBean. Each invocation requires the operation name to be invoked and the ObjectName of the target MBean. Both of these must be explicitly provided via adapter configuration:

<int-jmx:operation-invoking-channel-adapter id="adapter"
    object-name="example.domain:name=TestBean"
    operation-name="ping"/>

Then the adapter only needs to be able to discover the mbeanServer bean. If a different bean name is required, then provide the mbean-server attribute with a reference.

The payload of the Message will be mapped to the parameters of the operation, if any. A Map-typed payload with String keys is treated as name/value pairs, whereas a List or array would be passed as a simple argument list (with no explicit parameter names). If the operation requires a single parameter value, then the payload can represent that single value, and if the operation requires no parameters, then the payload would be ignored.

If you want to expose a channel for a single common operation to be invoked by Messages that need not contain headers, then that option works well.

Similar to the operation-invoking-channel-adapter Spring Integration also provides a operation-invoking-outbound-gateway, which could be used when dealing with non-void operations and a return value is required. Such return value will be sent as message payload to the reply-channel specified by this Gateway.

<int-jmx:operation-invoking-outbound-gateway request-channel="requestChannel"
   reply-channel="replyChannel"
   object-name="o.s.i.jmx.config:type=TestBean,name=testBeanGateway"
   operation-name="testWithReturn"/>

If the reply-channel attribute is not provided, the reply message will be sent to the channel that is identified by the IntegrationMessageHeaderAccessor.REPLY_CHANNEL header. That header is typically auto-created by the entry point into a message flow, such as any Gateway component. However, if the message flow was started by manually creating a Spring Integration Message and sending it directly to a Channel, then you must specify the message header explicitly or use the provided reply-channel attribute.

Spring Integration components themselves may be exposed as MBeans when the IntegrationMBeanExporter is configured. To create an instance of the IntegrationMBeanExporter, define a bean and provide a reference to an MBeanServer and a domain name (if desired). The domain can be left out, in which case the default domain is org.springframework.integration.

<int-jmx:mbean-export id="integrationMBeanExporter"
            default-domain="my.company.domain" server="mbeanServer"/>

<bean id="mbeanServer" class="org.springframework.jmx.support.MBeanServerFactoryBean">
    <property name="locateExistingServerIfPossible" value="true"/>
</bean>

Important

The MBean exporter is orthogonal to the one provided in Spring core - it registers message channels and message handlers, but not itself. You can expose the exporter itself, and certain other components in Spring Integration, using the standard <context:mbean-export/> tag. The exporter has a some metrics attached to it, for instance a count of the number of active handlers and the number of queued messages. It also has a useful operation, as discussed in the section called “Orderly Shutdown Managed Operation”.

Starting with Spring Integration 4.0 the @EnableIntegrationMBeanExport annotation has been introduced for convenient configuration of a default (integrationMbeanExporter) bean of type IntegrationMBeanExporter with several useful options at the @Configuration class level. For example:

@Configuration
@EnableIntegration
@EnableIntegrationMBeanExport(server = "mbeanServer", managedComponents = "input")
public class ContextConfiguration {

	@Bean
	public MBeanServerFactoryBean mbeanServer() {
		return new MBeanServerFactoryBean();
	}
}

If there is a need to provide more options, or have several IntegrationMBeanExporter beans e.g. for different MBean Servers, or to avoid conflicts with the standard Spring MBeanExporter (e.g. via @EnableMBeanExport), you can simply configure an IntegrationMBeanExporter as a generic bean.

MBean ObjectNames

All the MessageChannel, MessageHandler and MessageSource instances in the application are wrapped by the MBean exporter to provide management and monitoring features. The generated JMX object names for each component type are listed in the table below:

Table 9.3. MBean ObjectNames

Component Type	ObjectName
MessageChannel	o.s.i:type=MessageChannel,name=<channelName>
MessageSource	o.s.i:type=MessageSource,name=<channelName>,bean=<source>
MessageHandler	o.s.i:type=MessageSource,name=<channelName>,bean=<source>

The bean attribute in the object names for sources and handlers takes one of the values in the table below:

Table 9.4. bean ObjectName Part

Bean Value	Description
endpoint	The bean name of the enclosing endpoint (e.g. <service-activator>) if there is one
anonymous	An indication that the enclosing endpoint didn’t have a user-specified bean name, so the JMX name is the input channel name
internal	For well-known Spring Integration default components
handler/source	None of the above: fallback to the toString() of the object being monitored (handler or source)

Custom elements can be appended to the object name by providing a reference to a Properties object in the object-name-static-properties attribute.

Also, since Spring Integration 3.0, you can use a custom ObjectNamingStrategy using the object-naming-strategy attribute. This permits greater control over the naming of the MBeans. For example, to group all Integration MBeans under an Integration type. A simple custom naming strategy implementation might be:

public class Namer implements ObjectNamingStrategy {

	private final ObjectNamingStrategy realNamer = new KeyNamingStrategy();
	@Override
	public ObjectName getObjectName(Object managedBean, String beanKey) throws MalformedObjectNameException {
		String actualBeanKey = beanKey.replace("type=", "type=Integration,componentType=");
		return realNamer.getObjectName(managedBean, actualBeanKey);
	}

}

The beanKey argument is a String containing the standard object name beginning with the default-domain and including any additional static properties. This example simply moves the standard type part to componentType and sets the type to Integration, enabling selection of all Integration MBeans in one query:"my.domain:type=Integration,*. This also groups the beans under one tree entry under the domain in tools like VisualVM.

	Note
	The default naming strategy is a MetadataNamingStrategy. The exporter propagates the default-domain to that object to allow it to generate a fallback object name if parsing of the bean key fails. If your custom naming strategy is a MetadataNamingStrategy (or subclass), the exporter will not propagate the default-domain; you will need to configure it on your strategy bean.

JMX Improvements

Version 4.2 introduced some important improvements, representing a fairly major overhaul to the JMX support in the framework. These resulted in a significant performance improvement of the JMX statistics collection and much more control thereof, but has some implications for user code in a few specific (uncommon) situations. These changes are detailed below, with a caution where necessary.

• Metrics Capture

Previously, MessageSource, MessageChannel and MessageHandler metrics were captured by wrapping the object in a JDK dynamic proxy to intercept appropriate method calls and capture the statistics. The proxy was added when an integration MBean exporter was declared in the context.

Now, the statistics are captured by the beans themselves; see Section 9.1, “Metrics and Management” for more information.

Warning

This change means that you no longer automatically get an MBean or statistics for custom MessageHandler implementations, unless those custom handlers extend AbstractMessageHandler. The simplest way to resolve this is to extend AbstractMessageHandler. If that’s not possible, or desired, another work-around is to implement the MessageHandlerMetrics interface. For convenience, a DefaultMessageHandlerMetrics is provided to capture and report statistics. Invoke the beforeHandle and afterHandle at the appropriate times. Your MessageHandlerMetrics methods can then delegate to this object to obtain each statistic. Similarly, MessageSource implementations must extend AbstractMessageSource or implement MessageSourceMetrics. Message sources only capture a count so there is no provided convenience class; simply maintain the count in an AtomicLong field.

The removal of the proxy has two additional benefits; 1) stack traces in exceptions are reduced (when JMX is enabled) because the proxy is not on the stack; 2) cases where 2 MBeans were exported for the same bean now only export a single MBean with consolidated attributes/operations (see the MBean consolidation bullet below).

• Resolution

System.nanoTime() is now used to capture times instead of System.currentTimeMillis(). This may provide more accuracy on some JVMs, espcially when durations of less than 1 millisecond are expected

• Setting Initial Statistics Collection State

Previously, when JMX was enabled, all sources, channels, handlers captured statistics. It is now possible to control whether the statisics are enabled on an individual component. Further, it is possible to capture simple counts on MessageChannel s and MessageHandler s instead of the complete time-based statistics. This can have significant performance implications because you can selectively configure where you need detailed statistics, as well as enable/disable at runtime.

See Section 9.1, “Metrics and Management”.

• @IntegrationManagedResource

Similar to the @ManagedResource annotation, the @IntegrationManagedResource marks a class as eligible to be exported as an MBean; however, it will only be exported if there is an IntegrationMBeanExporter in the application context.

Certain Spring Integration classes (in the org.springframework.integration) package) that were previously annotated with`@ManagedResource` are now annotated with both @ManagedResource and @IntegrationManagedResource. This is for backwards compatibility (see the next bullet). Such MBeans will be exported by any context MBeanServeror an IntegrationMBeanExporter (but not both - if both exporters are present, the bean is exported by the integration exporter if the bean matches a managed-components pattern).

• Consolidated MBeans

Certain classes within the framework (mapping routers for example) have additional attributes/operations over and above those provided by metrics and Lifecycle. We will use a Router as an example here.

Previously, beans of these types were exported as two distinct MBeans:

1) the metrics MBean (with an objectName such as: intDomain:type=MessageHandler,name=myRouter,bean=endpoint). This MBean had metrics attributes and metrics/Lifecycle operations.

2) a second MBean (with an objectName such as: ctxDomain:name=org.springframework.integration.config.RouterFactoryBean#0 ,type=MethodInvokingRouter) was exported with the channel mappings attribute and operations.

Now, the attributes and operations are consolidated into a single MBean. The objectName will depend on the exporter. If exported by the integration MBean exporter, the objectName will be, for example: intDomain:type=MessageHandler,name=myRouter,bean=endpoint. If exported by another exporter, the objectName will be, for example: ctxDomain:name=org.springframework.integration.config.RouterFactoryBean#0 ,type=MethodInvokingRouter. There is no difference between these MBeans (aside from the objectName), except that the statistics will not be enabled (the attributes will be 0) by exporters other than the integration exporter; statistics can be enabled at runtime using the JMX operations. When exported by the integration MBean exporter, the initial state can be managed as described above.

	Warning
	If you are currently using the second MBean to change, for example, channel mappings, and you are using the integration MBean exporter, note that the objectName has changed because of the MBean consolidation. There is no change if you are not using the integration MBean exporter.

• MBean Exporter Bean Name Patterns

Previously, the managed-components patterns were inclusive only. If a bean name matched one of the patterns it would be included. Now, the pattern can be negated by prefixing it with !. i.e. "!foo*, foox" will match all beans that don’t start with foo, except foox. Patterns are evaluated left to right and the first match (positive or negative) wins and no further patterns are applied.

	Warning
	The addition of this syntax to the pattern causes one possible (although perhaps unlikey) problem. If you have a bean "!foo"and you included a pattern "!foo" in your MBean exporter’s managed-components patterns; it will no long match; the pattern will now match all beans not named foo. In this case, you can escape the ! in the pattern with \. The pattern "\!foo" means match a bean named "!foo".

• IntegrationMBeanExporter changes

The IntegrationMBeanExporter no longer implements SmartLifecycle; this means that start() and stop() operations are no longer available to register/unregister MBeans. The MBeans are now registered during context initialization and unregistered when the context is destroyed.

public void stopActiveComponents(long howLong)

To enable Message History all you need is to define the message-history element in your configuration.

<int:message-history/>

Now every named component (component that has an id defined) will be tracked. The framework will set the history header in your Message. Its value is very simple - List<Properties>.

<int:gateway id="sampleGateway" 
    service-interface="org.springframework.integration.history.sample.SampleGateway"
    default-request-channel="bridgeInChannel"/>

<int:chain id="sampleChain" input-channel="chainChannel" output-channel="filterChannel">
  <int:header-enricher>
    <int:header name="baz" value="baz"/>
  </int:header-enricher>
</int:chain>

The above configuration will produce a very simple Message History structure:

[{name=sampleGateway, type=gateway, timestamp=1283281668091},
 {name=sampleChain, type=chain, timestamp=1283281668094}]

To get access to Message History all you need is access the MessageHistory header. For example:

Iterator<Properties> historyIterator =
    message.getHeaders().get(MessageHistory.HEADER_NAME, MessageHistory.class).iterator();
assertTrue(historyIterator.hasNext());
Properties gatewayHistory = historyIterator.next();
assertEquals("sampleGateway", gatewayHistory.get("name"));
assertTrue(historyIterator.hasNext());
Properties chainHistory = historyIterator.next();
assertEquals("sampleChain", chainHistory.get("name"));

You might not want to track all of the components. To limit the history to certain components based on their names, all you need is provide the tracked-components attribute and specify a comma-delimited list of component names and/or patterns that match the components you want to track.

<int:message-history tracked-components="*Gateway, sample*, foo"/>

In the above example, Message History will only be maintained for all of the components that end with Gateway, start with sample, or match the name foo exactly.

Starting with version 4.0, you can also use the @EnableMessageHistory annotation in a @Configuration class. In addition, the MessageHistoryConfigurer bean is now exposed as a JMX MBean by the IntegrationMBeanExporter (see Section 9.2.7, “MBean Exporter”), allowing the patterns to be changed at runtime. Note, however, that the bean must be stopped (turning off message history) in order to change the patterns. This feature might be useful to temporarily turn on history to analyze a system. The MBean’s object name is "<domain>:name=messageHistoryConfigurer,type=MessageHistoryConfigurer".

	Important
	If multiple beans (declared by @EnableMessageHistory and/or <message-history/>) they all must have identical component name patterns (when trimmed and sorted). Do not use a generic <bean/> definition for the MessageHistoryConfigurer.

Note

Remember that by definition the Message History header is immutable (you can’t re-write history, although some try). Therefore, when writing Message History values, the components are either creating brand new Messages (when the component is an origin), or they are copying the history from a request Message, modifying it and setting the new list on a reply Message. In either case, the values can be appended even if the Message itself is crossing thread boundaries. That means that the history values can greatly simplify debugging in an asynchronous message flow.

Starting with version 4.3, some MessageGroupStore implementations can be injected with a custom MessageGroupFactory strategy to create/customize the MessageGroup instances used by the MessageGroupStore. This defaults to a SimpleMessageGroupFactory which produces SimpleMessageGroup s based on the GroupType.HASH_SET (LinkedHashSet) internal collection. Other possible options are SYNCHRONISED_SET and BLOCKING_QUEUE, where the last one can be used to reinstate the previous SimpleMessageGroup behavior. Also the PERSISTENT option is available. See the next section for more information.

Starting with version 4.3, all persistence MessageGroupStore s retrieve MessageGroup s and their messages from the store with the Lazy-Load manner. In most cases it is useful for the Correlation MessageHandler s (Section 6.4, “Aggregator” and Section 6.5, “Resequencer”), when it would be an overhead to load entire MessageGroup from the store on each correlation operation.

To switch off the lazy-load behavior the AbstractMessageGroupStore.setLazyLoadMessageGroups(false) option can be used from the configuration.

Our performance tests for lazy-load on MongoDB MessageStore (Section 22.3, “MongoDB Message Store”) and <aggregator> (Section 6.4, “Aggregator”) with custom release-strategy like:

<int:aggregator input-channel="inputChannel"
                output-channel="outputChannel"
                message-store="mongoStore"
                release-strategy-expression="size() == 1000"/>

demonstrate this results for 1000 simple messages:

StopWatch 'Lazy-Load Performance': running time (millis) = 38918
-----------------------------------------
ms     %     Task name
-----------------------------------------
02652  007%  Lazy-Load
36266  093%  Eager

The Metadata Store is useful for implementing the EIP Idempotent Receiver pattern, when there is need to filter an incoming Message if it has already been processed, and just discard it or perform some other logic on discarding. The following configuration is an example of how to do this:

<int:filter input-channel="serviceChannel"
			output-channel="idempotentServiceChannel"
			discard-channel="discardChannel"
			expression="@metadataStore.get(headers.businessKey) == null"/>

<int:publish-subscribe-channel id="idempotentServiceChannel"/>

<int:outbound-channel-adapter channel="idempotentServiceChannel"
                              expression="@metadataStore.put(headers.businessKey, '')"/>

<int:service-activator input-channel="idempotentServiceChannel" ref="service"/>

The value of the idempotent entry may be some expiration date, after which that entry should be removed from Metadata Store by some scheduled reaper.

Also see Section 8.9.9, “Idempotent Receiver Enterprise Integration Pattern”.

Some metadata stores (currently only zookeeper) support registering a listener to receive events when items change.

public interface MetadataStoreListener {

	void onAdd(String key, String value);

	void onRemove(String key, String oldValue);

	void onUpdate(String key, String newValue);
}

See the javadocs for more information. The MetadataStoreListenerAdapter can be subclassed if you are only interested in a subset of events.

Spring Integration components have various levels of complexity. For example, any polled MessageSource also has a SourcePollingChannelAdapter and a MessageChannel to which to send messages from the source data periodically. Other components might be middleware request-reply components, e.g. JmsOutboundGateway, with a consuming AbstractEndpoint to subscribe to (or poll) the requestChannel (input) for messages, and a replyChannel (output) to produce a reply message to send downstream. Meanwhile, any MessageProducerSupport implementation (e.g. ApplicationEventListeningMessageProducer) simply wraps some source protocol listening logic and sends messages to the outputChannel.

Within the graph, Spring Integration components are represented using the IntegrationNode class hierarchy, which you can find in the o.s.i.support.management.graph package. For example the ErrorCapableDiscardingMessageHandlerNode could be used for the AggregatingMessageHandler (because it has a discardChannel option) and can produce errors when consuming from a PollableChannel using a PollingConsumer. Another sample is CompositeMessageHandlerNode - for a MessageHandlerChain when subscribed to a SubscribableChannel, using an EventDrivenConsumer.

	Note
	The @MessagingGateway (see Section 8.4, “Messaging Gateways”) provides nodes for each its method, where the name attribute is based on the gateway’s bean name and the short method signature. For example the gateway:

@MessagingGateway(defaultRequestChannel = "four")
public interface Gate {

	void foo(String foo);

	void foo(Integer foo);

	void bar(String bar);

}

produces nodes like:

{
  "nodeId" : 10,
  "name" : "gate.bar(class java.lang.String)",
  "stats" : null,
  "componentType" : "gateway",
  "output" : "four",
  "errors" : null
},
{
  "nodeId" : 11,
  "name" : "gate.foo(class java.lang.String)",
  "stats" : null,
  "componentType" : "gateway",
  "output" : "four",
  "errors" : null
},
{
  "nodeId" : 12,
  "name" : "gate.foo(class java.lang.Integer)",
  "stats" : null,
  "componentType" : "gateway",
  "output" : "four",
  "errors" : null
}

This IntegrationNode hierarchy can be used for parsing the graph model on the client side, as well as for the understanding the general Spring Integration runtime behavior. See also Section 3.7, “Programming Tips and Tricks” for more information.