System Management

Metrics and Management

This section describes how to capture metrics for Spring Integration. In recent versions, we have relied more on Micrometer (see https://micrometer.io), and we plan to use Micrometer even more in future releases.

Disabling Logging in High Volume Environments

You can control debug logging in the main message flow. In very high volume applications, 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) is not affected by this setting.

The following listing shows the available options for controlling logging:

Java
@Configuration
@EnableIntegration
@EnableIntegrationManagement(
    defaultLoggingEnabled = "true" <1>)

public static class ContextConfiguration {
...
}
XML
<int:management default-logging-enabled="true"/> (1)
1 Set to false to disable all logging in the main message flow, regardless of the log system category settings. Set to 'true' to enable debug logging (if also enabled by the logging subsystem). Only applied if you have not explicitly configured the setting in a bean definition. The default is true.
defaultLoggingEnabled is applied only if you have not explicitly configured the corresponding setting in a bean definition.

Micrometer Integration

Overview

Starting with version 5.0.3, the presence of a Micrometer MeterRegistry in the application context triggers support for Micrometer metrics.

To use Micrometer, add one of the MeterRegistry beans to the application context.

For each MessageHandler and MessageChannel, timers are registered. For each MessageSource, a counter is registered.

This only applies to objects that extend AbstractMessageHandler, AbstractMessageChannel, and AbstractMessageSource (which is the case for most framework components).

The Timer Meters for send operations on message channels have the following names or tags:

  • name: spring.integration.send

  • tag: type:channel

  • tag: name:<componentName>

  • tag: result:(success|failure)

  • tag: exception:(none|exception simple class name)

  • description: Send processing time

(A failure result with a none exception means the channel’s send() operation returned false.)

The Counter Meters for receive operations on pollable message channels have the following names or tags:

  • name: spring.integration.receive

  • tag: type:channel

  • tag: name:<componentName>

  • tag: result:(success|failure)

  • tag: exception:(none|exception simple class name)

  • description: Messages received

The Timer Meters for operations on message handlers have the following names or tags:

  • name: spring.integration.send

  • tag: type:handler

  • tag: name:<componentName>

  • tag: result:(success|failure)

  • tag: exception:(none|exception simple class name)

  • description: Send processing time

The Counter meters for message sources have the following names/tags:

  • name: spring.integration.receive

  • tag: type:source

  • tag: name:<componentName>

  • tag: result:success

  • tag: exception:none

  • description: Messages received

In addition, there are three Gauge Meters:

  • spring.integration.channels: The number of MessageChannels in the application.

  • spring.integration.handlers: The number of MessageHandlers in the application.

  • spring.integration.sources: The number of MessageSources in the application.

It is possible to customize the names and tags of Meters created by integration components by providing a subclass of MicrometerMetricsCaptor. The MicrometerCustomMetricsTests test case shows a simple example of how to do that. You can also further customize the meters by overloading the build() methods on builder subclasses.

Starting with version 5.1.13, the QueueChannel exposes Micrometer gauges for queue size and remaining capacity:

  • name: spring.integration.channel.queue.size

  • tag: type:channel

  • tag: name:<componentName>

  • description: The size of the queue channel

and

  • name: spring.integration.channel.queue.remaining.capacity

  • tag: type:channel

  • tag: name:<componentName>

  • description: The remaining capacity of the queue channel

Disabling Meters

By default, all meters are registered when first used. Now, with Micrometer, you can add MeterFilter s to the MeterRegistry to prevent some or all from being registered. You can filter out (deny) meters by any of the properties provided, name, tag, etc. See Meter Filters in the Micrometer documentation for more information.

For example, given:

@Bean
public QueueChannel noMeters() {
    return new QueueChannel(10);
}

You can suppress registration of meters for just this channel with:

registry.config().meterFilter(MeterFilter.deny(id ->
        "channel".equals(id.getTag("type")) &&
        "noMeters".equals(id.getTag("name"))));

Micrometer Observation

Starting with version 6.0, Spring Integration utilizes a Micrometer Observation abstraction which can handle metrics as well as tracing via appropriate ObservationHandler configuration.

The observation handling is enabled on the IntegrationManagement components whenever an ObservationRegistry bean is present in the application context and an @EnableIntegrationManagement is configured. To customize what set of components should be instrumented, an observationPatterns() attribute is exposed on the @EnableIntegrationManagement annotation. See its javadocs for a pattern matching algorithm.

By default, none of the IntegrationManagement components are instrumented with an ObservationRegistry bean. Can be configured as * to match all components.

The meters are not gathered in this case independently, but delegated to an appropriate ObservationHandler configured on the provided ObservationRegistry.

The following Spring Integration components are instrumented with observation logic each with a respective convention:

  • MessageProducerSupport, being the inbound endpoint of the flow, is considered as a CONSUMER span type and uses the IntegrationObservation.HANDLER API;

  • MessagingGatewaySupport` is an inbound request-reply endpoint, and is considered as a SERVER span type. It uses the IntegrationObservation.GATEWAY API;

  • An AbstractMessageChannel.send() operation is the only Spring Integration API where it produces messages. So, it is treated as a PRODUCER span type and uses the IntegrationObservation.PRODCUER API. This makes more sense when a channel is a distributed implementation (e.g. PublishSubscribeKafkaChannel or ZeroMqChannel) and trace information has to be added to the message. So, the IntegrationObservation.PRODUCER observation is based on a MessageSenderContext where Spring Integration supplies a MutableMessage to allow a subsequent tracing Propagator to add headers, so they are available to the consumer;

  • An AbstractMessageHandler is a CONSUMER span type and uses the IntegrationObservation.HANDLER API.

An observation production on the IntegrationManagement components can be customized via ObservationConvention configuration. For example an AbstractMessageHandler expects a MessageReceiverObservationConvention via its setObservationConvention() API.

The following are supported metrics, spans and conventions for Observation API:

Observability - Metrics

Below you can find a list of all metrics declared by this project.

Gateway

Observation for inbound message gateways.

Metric name spring.integration.gateway (defined by convention class o.s.i.support.management.observation.DefaultMessageRequestReplyReceiverObservationConvention). Type timer.

Metric name spring.integration.gateway.active (defined by convention class o.s.i.support.management.observation.DefaultMessageRequestReplyReceiverObservationConvention). Type long task timer.

KeyValues that are added after starting the Observation might be missing from the *.active metrics.
Micrometer internally uses nanoseconds for the baseunit. However, each backend determines the actual baseunit. (i.e. Prometheus uses seconds)

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 1. Low cardinality Keys

Name

Description

spring.integration.name (required)

Name of the message gateway component.

spring.integration.outcome (required)

Outcome of the request/reply execution.

spring.integration.type (required)

Type of the component - 'gateway'.

Handler

Observation for message handlers.

Metric name spring.integration.handler (defined by convention class o.s.i.support.management.observation.DefaultMessageReceiverObservationConvention). Type timer.

Metric name spring.integration.handler.active (defined by convention class o.s.i.support.management.observation.DefaultMessageReceiverObservationConvention). Type long task timer.

KeyValues that are added after starting the Observation might be missing from the *.active metrics.
Micrometer internally uses nanoseconds for the baseunit. However, each backend determines the actual baseunit. (i.e. Prometheus uses seconds)

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 2. Low cardinality Keys

Name

Description

spring.integration.name (required)

Name of the message handler component.

spring.integration.type (required)

Type of the component - 'handler'.

Producer

Observation for message producers, e.g. channels.

Metric name spring.integration.producer (defined by convention class o.s.i.support.management.observation.DefaultMessageSenderObservationConvention). Type timer.

Metric name spring.integration.producer.active (defined by convention class o.s.i.support.management.observation.DefaultMessageSenderObservationConvention). Type long task timer.

KeyValues that are added after starting the Observation might be missing from the *.active metrics.
Micrometer internally uses nanoseconds for the baseunit. However, each backend determines the actual baseunit. (i.e. Prometheus uses seconds)

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 3. Low cardinality Keys

Name

Description

spring.integration.name (required)

Name of the message handler component.

spring.integration.type (required)

Type of the component - 'producer'.

Observability - Spans

Below you can find a list of all spans declared by this project.

Gateway Span

Observation for inbound message gateways.

Span name spring.integration.gateway (defined by convention class o.s.i.support.management.observation.DefaultMessageRequestReplyReceiverObservationConvention).

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 4. Tag Keys

Name

Description

spring.integration.name (required)

Name of the message gateway component.

spring.integration.outcome (required)

Outcome of the request/reply execution.

spring.integration.type (required)

Type of the component - 'gateway'.

Handler Span

Observation for message handlers.

Span name spring.integration.handler (defined by convention class o.s.i.support.management.observation.DefaultMessageReceiverObservationConvention).

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 5. Tag Keys

Name

Description

spring.integration.name (required)

Name of the message handler component.

spring.integration.type (required)

Type of the component - 'handler'.

Producer Span

Observation for message producers, e.g. channels.

Span name spring.integration.producer (defined by convention class o.s.i.support.management.observation.DefaultMessageSenderObservationConvention).

Fully qualified name of the enclosing class o.s.i.support.management.observation.IntegrationObservation.

All tags must be prefixed with spring.integration. prefix!
Table 6. Tag Keys

Name

Description

spring.integration.name (required)

Name of the message handler component.

spring.integration.type (required)

Type of the component - 'producer'.

Observability - Conventions

Below you can find a list of all GlobalObservationConvention and ObservationConvention declared by this project.

Table 7. ObservationConvention implementations

ObservationConvention Class Name

Applicable ObservationContext Class Name

o.s.i.support.management.observation.DefaultMessageReceiverObservationConvention

MessageReceiverContext

o.s.i.support.management.observation.MessageReceiverObservationConvention

MessageReceiverContext

o.s.i.support.management.observation.DefaultMessageRequestReplyReceiverObservationConvention

MessageRequestReplyReceiverContext

o.s.i.support.management.observation.MessageRequestReplyReceiverObservationConvention

MessageRequestReplyReceiverContext

o.s.i.support.management.observation.DefaultMessageSenderObservationConvention

MessageSenderContext

o.s.i.support.management.observation.MessageSenderObservationConvention

MessageSenderContext

Observation Propagation

To supply a connected chain of spans in one trace, independently of the nature of the messaging flow, even if a MessageChannel is persistent and distributed, the observation must be enabled on this channel and on consumers (subscribers) for this channel. This way, the tracing information is stored in the message headers before it is propagated to a consumer thread or persisted into the database. This is done via mentioned above MessageSenderContext. The consumer (a MessageHandler) side restores tracing information from those headers using a MessageReceiverContext and starts a new child Observation.

Spring Integration JMX Support

Also see JMX Support.

Message History

The key benefit of a messaging architecture is loose coupling such that participating components do not maintain any awareness about one another. This fact alone makes an application extremely flexible, letting you change components without affecting the rest of the flow, change messaging routes, change message consuming styles (polling versus event driven), and so on. However, this unassuming style of architecture could prove to be difficult when things go wrong. When debugging, you probably want as much information (its origin, the channels it has traversed, and other details) about the message as you can get.

Message history is one of those patterns that helps by giving you an option to maintain some level of awareness of a message path either for debugging purposes or for maintaining an audit trail. Spring integration provides a simple way to configure your message flows to maintain the message history by adding a header to the message and updating that header every time a message passes through a tracked component.

Message History Configuration

To enable message history, you need only define the message-history element (or @EnableMessageHistory) in your configuration, as shown in the following example:

Java
@Configuration
@EnableIntegration
@EnableMessageHistory
XML
<int:message-history/>

Now every named component (that has an 'id' defined) is tracked. The framework sets the 'history' header in your message. Its value a List<Properties>.

Consider the following configuration example:

Java
@MessagingGateway(defaultRequestChannel = "bridgeInChannel")
public interface SampleGateway {
   ...
}

@Bean
@Transformer(inputChannel = "enricherChannel", outputChannel="filterChannel")
HeaderEnricher sampleEnricher() {
    HeaderEnricher enricher =
           new HeaderEnricher(Collections.singletonMap("baz", new StaticHeaderValueMessageProcessor("baz")));
    return enricher;
}
XML
<int:gateway id="sampleGateway"
    service-interface="org.springframework.integration.history.sample.SampleGateway"
    default-request-channel="bridgeInChannel"/>

<int:header-enricher id="sampleEnricher" input-channel="enricherChannel" output-channel="filterChannel">
    <int:header name="baz" value="baz"/>
</int:header-enricher>

The preceding configuration produces a simple message history structure, with output similar to the following:

[{name=sampleGateway, type=gateway, timestamp=1283281668091},
 {name=sampleEnricher, type=header-enricher, timestamp=1283281668094}]

To get access to message history, you need only access the MessageHistory header. The following example shows how to do so:

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 the components. To limit the history to certain components based on their names, you can provide the tracked-components attribute and specify a comma-delimited list of component names and patterns that match the components you want to track. The following example shows how to do so:

Java
@Configuration
@EnableIntegration
@EnableMessageHistory("*Gateway", "sample*", "aName")
XML
<int:message-history tracked-components="*Gateway, sample*, aName"/>

In the preceding example, message history is maintained only for the components that end with 'Gateway', start with 'sample', or match the name, 'aName', exactly.

In addition, the MessageHistoryConfigurer bean is now exposed as a JMX MBean by the IntegrationMBeanExporter (see MBean Exporter), letting you change the patterns 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.

Only one @EnableMessageHistory (or <message-history/>) must be declared in the application context as single source for components tracking configuration. Do not use a generic bean definition for the MessageHistoryConfigurer.
By definition, the message history header is immutable (you cannot re-write history). Therefore, when writing message history values, the components either create new messages (when the component is an origin) or they copy 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.

Message Store

The Enterprise Integration Patterns (EIP) book identifies several patterns that have the ability to buffer messages. For example, an aggregator buffers messages until they can be released, and a QueueChannel buffers messages until consumers explicitly receive those messages from that channel. Because of the failures that can occur at any point within your message flow, EIP components that buffer messages also introduce a point where messages could be lost.

To mitigate the risk of losing messages, EIP defines the message store pattern, which lets EIP components store messages, typically in some type of persistent store (such as an RDBMS).

Spring Integration provides support for the message store pattern by:

  • Defining an org.springframework.integration.store.MessageStore strategy interface

  • Providing several implementations of this interface

  • Exposing a message-store attribute on all components that have the capability to buffer messages so that you can inject any instance that implements the MessageStore interface.

Details on how to configure a specific message store implementation and how to inject a MessageStore implementation into a specific buffering component are described throughout the manual (see the specific component, such as QueueChannel, Aggregator, Delayer, and others). The following pair of examples show how to add a reference to a message store for a QueueChannel and for an aggregator:

Example 1. QueueChannel
<int:channel id="myQueueChannel">
    <int:queue message-store="refToMessageStore"/>
<int:channel>
Example 2. Aggregator
<int:aggregator … message-store="refToMessageStore"/>

By default, messages are stored in-memory by using o.s.i.store.SimpleMessageStore, an implementation of MessageStore. That might be fine for development or simple low-volume environments where the potential loss of non-persistent messages is not a concern. However, the typical production application needs a more robust option, not only to mitigate the risk of message loss but also to avoid potential out-of-memory errors. Therefore, we also provide MessageStore implementations for a variety of data-stores. The following is a complete list of supported implementations:

However, be aware of some limitations while using persistent implementations of the MessageStore.

The Message data (payload and headers) is serialized and deserialized by using different serialization strategies, depending on the implementation of the MessageStore. For example, when using JdbcMessageStore, only Serializable data is persisted by default. In this case, non-Serializable headers are removed before serialization occurs. Also, be aware of the protocol-specific headers that are injected by transport adapters (such as FTP, HTTP, JMS, and others). For example, <http:inbound-channel-adapter/> maps HTTP headers into message headers, and one of them is an ArrayList of non-serializable org.springframework.http.MediaType instances. However, you can inject your own implementation of the Serializer and Deserializer strategy interfaces into some MessageStore implementations (such as JdbcMessageStore) to change the behavior of serialization and deserialization.

Pay special attention to the headers that represent certain types of data. For example, if one of the headers contains an instance of some Spring bean, upon deserialization, you may end up with a different instance of that bean, which directly affects some of the implicit headers created by the framework (such as REPLY_CHANNEL or ERROR_CHANNEL). Currently, they are not serializable, but, even if they were, the deserialized channel would not represent the expected instance.

Beginning with Spring Integration version 3.0, you can resolve this issue with a header enricher configured to replace these headers with a name after registering the channel with the HeaderChannelRegistry.

Also, consider what happens when you configure a message-flow as follows: gateway → queue-channel (backed by a persistent Message Store) → service-activator. That gateway creates a temporary reply channel, which is lost by the time the service-activator’s poller reads from the queue. Again, you can use the header enricher to replace the headers with a String representation.

For more information, see Header Enricher.

Spring Integration 4.0 introduced two new interfaces:

  • ChannelMessageStore: To implement operations specific for QueueChannel instances

  • PriorityCapableChannelMessageStore: To mark MessageStore implementations to be used for PriorityChannel instances and to provide priority order for persisted messages.

The real behavior depends on the implementation. The framework provides the following implementations, which can be used as a persistent MessageStore for QueueChannel and PriorityChannel:

Caution about SimpleMessageStore

Starting with version 4.1, the SimpleMessageStore no longer copies the message group when calling getMessageGroup(). For large message groups, this was a significant performance problem. 4.0.1 introduced a boolean copyOnGet property that lets you control this behavior. When used internally by the aggregator, this property was set to false to improve performance. It is now false by default.

Users accessing the group store outside of components such as aggregators now get a direct reference to the group being used by the aggregator instead of a copy. Manipulation of the group outside the aggregator may cause unpredictable results.

For this reason, you should either not perform such manipulation or set the copyOnGet property to true.

Using MessageGroupFactory

Starting with version 4.3, some MessageGroupStore implementations can be injected with a custom MessageGroupFactory strategy to create and customize the MessageGroup instances used by the MessageGroupStore. This defaults to a SimpleMessageGroupFactory, which produces SimpleMessageGroup instances 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 5.0.1, the LIST option is also available for when the order and uniqueness of messages in the group does not matter.

Persistent MessageGroupStore and Lazy-load

Starting with version 4.3, all persistent MessageGroupStore instances retrieve MessageGroup instances and their messages from the store in the lazy-load manner. In most cases, it is useful for the correlation MessageHandler instances (see Aggregator and Resequencer), when it would add overhead to load entire the MessageGroup from the store on each correlation operation.

You can use the AbstractMessageGroupStore.setLazyLoadMessageGroups(false) option to switch off the lazy-load behavior from the configuration.

Our performance tests for lazy-load on MongoDB MessageStore (MongoDB Message Store) and <aggregator> (Aggregator) use a custom release-strategy similar to the following:

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

It produces results similar to the following for 1000 simple messages:

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

However, starting with version 5.5, all the persistent MessageGroupStore implementations provide a streamMessagesForGroup(Object groupId) contract based on the target database streaming API. This improves resources utilization when groups are very big in the store. Internally in the framework this new API is used in the Delayer (for example) when it reschedules persisted messages on startup. A returned Stream<Message<?>> must be closed in the end of processing, e.g. via auto-close by the try-with-resources. Whenever a PersistentMessageGroup is used, its streamMessages() delegates to the MessageGroupStore.streamMessagesForGroup().

Message Group Condition

Starting with version 5.5, the MessageGroup abstraction provides a condition string option. The value of this option can be anything that could be parsed later on for any reason to make a decision for the group. For example a ReleaseStrategy from a correlation message handler may consult this property from the group instead of iterating all the messages in the group. The MessageGroupStore exposes a setGroupCondition(Object groupId, String condition) API. For this purpose a setGroupConditionSupplier(BiFunction<Message<?>, String, String>) option has been added to the AbstractCorrelatingMessageHandler. This function is evaluated against each message after it has been added to the group as well as the existing condition of the group. The implementation may decide to return a new value, the existing value, or reset the target condition to null. The value for a condition can be a JSON, SpEL expression, number or anything what can be serialized as a string and parsed afterwards. For example, the FileMarkerReleaseStrategy from the File Aggregator component, populates a condition into a group from the FileHeaders.LINE_COUNT header of the FileSplitter.FileMarker.Mark.END message and consults with it from its canRelease() comparing a group size with the value in this condition. This way it doesn’t iterate all the messages in group to find a FileSplitter.FileMarker.Mark.END message with the FileHeaders.LINE_COUNT header. It also allows the end marker to arrive at the aggregator before all the other records; for example when processing a file in a multi-threaded environment.

In addition, for configuration convenience, a GroupConditionProvider contract has been introduced. The AbstractCorrelatingMessageHandler checks if the provided ReleaseStrategy implements this interface and extracts a conditionSupplier for group condition evaluation logic.

Metadata Store

Many external systems, services, or resources are not transactional (Twitter, RSS, file systems, and so on), and there is not any ability to mark the data as read. Also, sometimes, you may need to implement the Enterprise Integration Pattern idempotent receiver in some integration solutions. To achieve this goal and store some previous state of the endpoint before the next interaction with external system or to deal with the next message, Spring Integration provides the metadata store component as an an implementation of the org.springframework.integration.metadata.MetadataStore interface with a general key-value contract.

The metadata store is designed to store various types of generic metadata (for example, the published date of the last feed entry that has been processed) to help components such as the feed adapter deal with duplicates. If a component is not directly provided with a reference to a MetadataStore, the algorithm for locating a metadata store is as follows: First, look for a bean with a metadataStore ID in the application context. If one is found, use it. Otherwise, create a new instance of SimpleMetadataStore, which is an in-memory implementation that persists only metadata within the lifecycle of the currently running application context. This means that, upon restart, you may end up with duplicate entries.

If you need to persist metadata between application context restarts, the framework provides the following persistent MetadataStores:

The PropertiesPersistingMetadataStore is backed by a properties file and a PropertiesPersister.

By default, it persists only the state when the application context is closed normally. It implements Flushable so that you can persist the state at will, by invoking flush(). The following example shows how to configure a 'PropertiesPersistingMetadataStore' with XML:

<bean id="metadataStore"
    class="org.springframework.integration.metadata.PropertiesPersistingMetadataStore"/>

Alternatively, you can provide your own implementation of the MetadataStore interface (for example, JdbcMetadataStore) and configure it as a bean in the application context.

Starting with version 4.0, SimpleMetadataStore, PropertiesPersistingMetadataStore, and RedisMetadataStore implement ConcurrentMetadataStore. These provide for atomic updates and can be used across multiple component or application instances.

Idempotent Receiver and Metadata Store

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 you can discard it or perform some other logic on discarding. The following configuration shows an example of how to do so:

<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 an expiration date, after which that entry should be removed from metadata store by some scheduled reaper.

MetadataStoreListener

Some metadata stores (currently only zookeeper) support registering a listener to receive events when items change, as the following example shows:

public interface MetadataStoreListener {

	void onAdd(String key, String value);

	void onRemove(String key, String oldValue);

	void onUpdate(String key, String newValue);
}

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

Control Bus

As described in the Enterprise Integration Patterns (EIP) book, the idea behind the control bus is that the same messaging system can be used for monitoring and managing the components within the framework as is used for “application-level” messaging. In Spring Integration, we build upon the adapters described above so that you can send messages as a means of invoking exposed operations.

The following example shows how to configure a control bus with XML:

<int:control-bus input-channel="operationChannel"/>

The control bus has an input channel that can be accessed for invoking operations on the beans in the application context. It also has all the common properties of a service activating endpoint. For example, you can specify an output channel if the result of the operation has a return value that you want to send on to a downstream channel.

The control bus runs messages on the input channel as Spring Expression Language (SpEL) expressions. It takes a message, compiles the body to an expression, adds some context, and then runs it. The default context supports any method that has been annotated with @ManagedAttribute or @ManagedOperation. It also supports the methods on Spring’s Lifecycle interface (and its Pausable extension since version 5.2), and it supports methods that are used to configure several of Spring’s TaskExecutor and TaskScheduler implementations. The simplest way to ensure that your own methods are available to the control bus is to use the @ManagedAttribute or @ManagedOperation annotations. Since those annotations are also used for exposing methods to a JMX MBean registry, they offer a convenient by-product: Often, the same types of operations you want to expose to the control bus are reasonable for exposing through JMX). Resolution of any particular instance within the application context is achieved in the typical SpEL syntax. To do so, provide the bean name with the SpEL prefix for beans (@). For example, to execute a method on a Spring Bean, a client could send a message to the operation channel as follows:

Message operation = MessageBuilder.withPayload("@myServiceBean.shutdown()").build();
operationChannel.send(operation)

The root of the context for the expression is the Message itself, so you also have access to the payload and headers as variables within your expression. This is consistent with all the other expression support in Spring Integration endpoints.

With Java annotations, you can configured the control bus as follows:

@Bean
@ServiceActivator(inputChannel = "operationChannel")
public ExpressionControlBusFactoryBean controlBus() {
    return new ExpressionControlBusFactoryBean();
}

Similarly, you can configure Java DSL flow definitions as follows:

@Bean
public IntegrationFlow controlBusFlow() {
    return IntegrationFlow.from("controlBus")
              .controlBus()
              .get();
}

If you prefer to use lambdas with automatic DirectChannel creation, you can create a control bus as follows:

@Bean
public IntegrationFlow controlBus() {
    return IntegrationFlowDefinition::controlBus;
}

In this case, the channel is named controlBus.input.

Orderly Shutdown

As described in "MBean Exporter", the MBean exporter provides a JMX operation called stopActiveComponents, which is used to stop the application in an orderly manner. The operation has a single Long parameter. The parameter indicates how long (in milliseconds) the operation waits to allow in-flight messages to complete. The operation works as follows:

  1. Call beforeShutdown() on all beans that implement OrderlyShutdownCapable.

    Doing so lets such components prepare for shutdown. Examples of components that implement this interface and what they do with this call include JMS and AMQP message-driven adapters that stop their listener containers, TCP server connection factories that stop accepting new connections (while keeping existing connections open), TCP inbound endpoints that drop (log) any new messages received, and HTTP inbound endpoints that return 503 - Service Unavailable for any new requests.

  2. Stop any active channels, such as JMS- or AMQP-backed channels.

  3. Stop all MessageSource instances.

  4. Stop all inbound MessageProducer s (that are not OrderlyShutdownCapable).

  5. Wait for any remaining time left, as defined by the value of the Long parameter passed in to the operation.

    Doing so lets any in-flight messages complete their journeys. It is therefore important to select an appropriate timeout when invoking this operation.

  6. Call afterShutdown() on all OrderlyShutdownCapable components.

    Doing so lets such components perform final shutdown tasks (closing all open sockets, for example).

As discussed in Orderly Shutdown Managed Operation, this operation can be invoked by using JMX. If you wish to programmatically invoke the method, you need to inject or otherwise get a reference to the IntegrationMBeanExporter. If no id attribute is provided on the <int-jmx:mbean-export/> definition, the bean has a generated name. This name contains a random component to avoid ObjectName collisions if multiple Spring Integration contexts exist in the same JVM (MBeanServer).

For this reason, if you wish to invoke the method programmatically, we recommend that you provide the exporter with an id attribute so that you can easily access it in the application context.

Finally, the operation can be invoked by using the <control-bus> element. See the monitoring Spring Integration sample application for details.

The algorithm described earlier was improved in version 4.1. Previously, all task executors and schedulers were stopped. This could cause mid-flow messages in QueueChannel instances to remain. Now the shutdown leaves pollers running, to let these messages be drained and processed.

Integration Graph

Starting with version 4.3, Spring Integration provides access to an application’s runtime object model, which can, optionally, include component metrics. It is exposed as a graph, which may be used to visualize the current state of the integration application. The o.s.i.support.management.graph package contains all the required classes to collect, build, and render the runtime state of Spring Integration components as a single tree-like Graph object. The IntegrationGraphServer should be declared as a bean to build, retrieve, and refresh the Graph object. The resulting Graph object can be serialized to any format, although JSON is flexible and convenient to parse and represent on the client side. A Spring Integration application with only the default components would expose a graph as follows:

{
  "contentDescriptor" : {
    "providerVersion" : "6.1.9",
    "providerFormatVersion" : 1.2,
    "provider" : "spring-integration",
    "name" : "myAppName:1.0"
  },
  "nodes" : [ {
    "nodeId" : 1,
    "componentType" : "null-channel",
    "integrationPatternType" : "null_channel",
    "integrationPatternCategory" : "messaging_channel",
    "properties" : { },
    "sendTimers" : {
      "successes" : {
        "count" : 1,
        "mean" : 0.0,
        "max" : 0.0
      },
      "failures" : {
        "count" : 0,
        "mean" : 0.0,
        "max" : 0.0
      }
    },
    "receiveCounters" : {
      "successes" : 0,
      "failures" : 0
    },
    "name" : "nullChannel"
  }, {
    "nodeId" : 2,
    "componentType" : "publish-subscribe-channel",
    "integrationPatternType" : "publish_subscribe_channel",
    "integrationPatternCategory" : "messaging_channel",
    "properties" : { },
    "sendTimers" : {
      "successes" : {
        "count" : 1,
        "mean" : 7.807002,
        "max" : 7.807002
      },
      "failures" : {
        "count" : 0,
        "mean" : 0.0,
        "max" : 0.0
      }
    },
    "name" : "errorChannel"
  }, {
    "nodeId" : 3,
    "componentType" : "logging-channel-adapter",
    "integrationPatternType" : "outbound_channel_adapter",
    "integrationPatternCategory" : "messaging_endpoint",
    "properties" : { },
    "output" : null,
    "input" : "errorChannel",
    "sendTimers" : {
      "successes" : {
        "count" : 1,
        "mean" : 6.742722,
        "max" : 6.742722
      },
      "failures" : {
        "count" : 0,
        "mean" : 0.0,
        "max" : 0.0
      }
    },
    "name" : "errorLogger"
  } ],
  "links" : [ {
    "from" : 2,
    "to" : 3,
    "type" : "input"
  } ]
}
Version 5.2 deprecated the legacy metrics in favor of Micrometer meters as discussed Metrics Management. The legacy metrics were removed in Version 5.4 and will no longer appear in the graph.

In the preceding example, the graph consists of three top-level elements.

The contentDescriptor graph element contains general information about the application providing the data. The name can be customized on the IntegrationGraphServer bean or in the spring.application.name application context environment property. Other properties are provided by the framework and let you distinguish a similar model from other sources.

The links graph element represents connections between nodes from the nodes graph element and, therefore, between integration components in the source Spring Integration application. For example, from a MessageChannel to an EventDrivenConsumer with some MessageHandler or from an AbstractReplyProducingMessageHandler to a MessageChannel. For convenience and to let you determine a link’s purpose, the model includes the type attribute. The possible types are:

  • input: Identifies the direction from MessageChannel to the endpoint, inputChannel, or requestChannel property

  • output: The direction from the MessageHandler, MessageProducer, or SourcePollingChannelAdapter to the MessageChannel through an outputChannel or replyChannel property

  • error: From MessageHandler on PollingConsumer or MessageProducer or SourcePollingChannelAdapter to the MessageChannel through an errorChannel property;

  • discard: From DiscardingMessageHandler (such as MessageFilter) to the MessageChannel through an errorChannel property.

  • route: From AbstractMappingMessageRouter (such as HeaderValueRouter) to the MessageChannel. Similar to output but determined at run-time. Maybe a configured channel mapping or a dynamically resolved channel. Routers typically retain only up to 100 dynamic routes for this purpose, but you can modify this value by setting the dynamicChannelLimit property.

The information from this element can be used by a visualization tool to render connections between nodes from the nodes graph element, where the from and to numbers represent the value from the nodeId property of the linked nodes. For example, the link element can be used to determine the proper port on the target node.

The following “text image” shows the relationships between the types:

              +---(discard)
              |
         +----o----+
         |         |
         |         |
         |         |
(input)--o         o---(output)
         |         |
         |         |
         |         |
         +----o----+
              |
              +---(error)

The nodes graph element is perhaps the most interesting, because its elements contain not only the runtime components with their componentType instances and name values but can also optionally contain metrics exposed by the component. Node elements contain various properties that are generally self-explanatory. For example, expression-based components include the expression property that contains the primary expression string for the component. To enable the metrics, add an @EnableIntegrationManagement to a @Configuration class or add an <int:management/> element to your XML configuration. See Metrics and Management for complete information.

The nodeId represents a unique incremental identifier to let you distinguish one component from another. It is also used in the links element to represent a relationship (connection) of this component to others, if any. The input and output attributes are for the inputChannel and outputChannel properties of the AbstractEndpoint, MessageHandler, SourcePollingChannelAdapter, or MessageProducerSupport. See the next section for more information.

Starting with version 5.1, the IntegrationGraphServer accepts a Function<NamedComponent, Map<String, Object>> additionalPropertiesCallback for population of additional properties on the IntegrationNode for a particular NamedComponent. For example, you can expose the SmartLifecycle autoStartup and running properties into the target graph:

server.setAdditionalPropertiesCallback(namedComponent -> {
            Map<String, Object> properties = null;
            if (namedComponent instanceof SmartLifecycle) {
                SmartLifecycle smartLifecycle = (SmartLifecycle) namedComponent;
                properties = new HashMap<>();
                properties.put("auto-startup", smartLifecycle.isAutoStartup());
                properties.put("running", smartLifecycle.isRunning());
            }
            return properties;
        });

Graph Runtime Model

Spring Integration components have various levels of complexity. For example, any polled MessageSource also has a SourcePollingChannelAdapter and a MessageChannel to which to periodically send messages from the source data. Other components might be middleware request-reply components (such as 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 (such as ApplicationEventListeningMessageProducer) wraps some source protocol listening logic and sends messages to the outputChannel.

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

The @MessagingGateway (see Messaging Gateways) provides nodes for each of its method, where the name attribute is based on the gateway’s bean name and the short method signature. Consider the following example of a gateway:
@MessagingGateway(defaultRequestChannel = "four")
public interface Gate {

	void foo(String foo);

	void foo(Integer foo);

	void bar(String bar);

}

The preceding gateway produces nodes similar to the following:

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

You can use this IntegrationNode hierarchy for parsing the graph model on the client side as well as to understand the general Spring Integration runtime behavior. See also Programming Tips and Tricks for more information.

Version 5.3 introduced an IntegrationPattern abstraction and all out-of-the-box components, which represent an Enterprise Integration Pattern (EIP), implement this abstraction and provide an IntegrationPatternType enum value. This information can be useful for some categorizing logic in the target application or, being exposed into the graph node, it can be used by a UI to determine how to draw the component.

Integration Graph Controller

If your application is web-based (or built on top of Spring Boot with an embedded web container) and the Spring Integration HTTP or WebFlux module (see HTTP Support and WebFlux Support, respectively) is present on the classpath, you can use a IntegrationGraphController to expose the IntegrationGraphServer functionality as a REST service. For this purpose, the @EnableIntegrationGraphController and @Configuration class annotations and the <int-http:graph-controller/> XML element are available in the HTTP module. Together with the @EnableWebMvc annotation (or <mvc:annotation-driven/> for XML definitions), this configuration registers an IntegrationGraphController @RestController where its @RequestMapping.path can be configured on the @EnableIntegrationGraphController annotation or <int-http:graph-controller/> element. The default path is /integration.

The IntegrationGraphController @RestController provides the following services:

  • @GetMapping(name = "getGraph"): To retrieve the state of the Spring Integration components since the last IntegrationGraphServer refresh. The o.s.i.support.management.graph.Graph is returned as a @ResponseBody of the REST service.

  • @GetMapping(path = "/refresh", name = "refreshGraph"): To refresh the current Graph for the actual runtime state and return it as a REST response. It is not necessary to refresh the graph for metrics. They are provided in real-time when the graph is retrieved. Refresh can be called if the application context has been modified since the graph was last retrieved. In that case, the graph is completely rebuilt.

You can set security and cross-origin restrictions for the IntegrationGraphController with the standard configuration options and components provided by the Spring Security and Spring MVC projects. The following example achieves those goals:

<mvc:annotation-driven />

<mvc:cors>
	<mvc:mapping path="/myIntegration/**"
				 allowed-origins="http://localhost:9090"
				 allowed-methods="GET" />
</mvc:cors>

<security:http>
    <security:intercept-url pattern="/myIntegration/**" access="ROLE_ADMIN" />
</security:http>


<int-http:graph-controller path="/myIntegration" />

The following example shows how to do the same thing with Java configuration:

@Configuration
@EnableWebMvc // or @EnableWebFlux
@EnableWebSecurity // or @EnableWebFluxSecurity
@EnableIntegration
@EnableIntegrationGraphController(path = "/testIntegration", allowedOrigins="http://localhost:9090")
public class IntegrationConfiguration extends WebSecurityConfigurerAdapter {

    @Override
    protected void configure(HttpSecurity http) throws Exception {
	    http
            .authorizeRequests()
               .antMatchers("/testIntegration/**").hasRole("ADMIN")
            // ...
            .formLogin();
    }

    //...

}

Note that, for convenience, the @EnableIntegrationGraphController annotation provides an allowedOrigins attribute. This provides GET access to the path. For more sophistication, you can configure the CORS mappings by using standard Spring MVC mechanisms.