Spring Boot AutoConfiguration

Axon's support for Spring Boot AutoConfiguration is by far the easiest option to get started configuring your Axon infrastructure components. By simply adding the axon-spring-boot-starter dependency, Axon will automatically configure the basic infrastructure components (Command Bus, Event Bus), as well as any component required to run and store Aggregates and Sagas.
Depending on other components available in the application context, Axon will define certain components, if they aren't already explicitly defined in the application context. This means that you only need to configure components that you want different from the default.
Event Bus and Event Store Configuration
If JPA is available, an Event Store with a JPA Event Storage Engine is used by default. This allow storage of Aggregates using Event Sourcing without any explicit configuration.
If JPA is not available, Axon defaults to a SimpleEventBus, which means that you need to specify a non-event sourcing repository for each Aggregate, or configure an EventStorageEngine in your Spring Configuration.
To configure a different Event Storage Engine, even if JPA is on the class path, simply define a bean of type EventStorageEngine (to use Event Sourcing) or EventBus (if Event Sourcing isn't required).
Command Bus Configuration
Axon will configure a SimpleCommandBus if no CommandBus implementation is explicitly defined in the Application Context. This CommandBus will use the TransactionManager to manage transactions.
If the only CommandBus bean defined is a DistributedCommandBus implementation, Axon will still configure a CommandBus implementation to serve as the local segment for the DistributedCommandBus. This bean will get a Qualifier "localSegment". It is recommended to define the DistributedCommandBus as a @Primary, so that it gets priority for dependency injection.
Query Bus Configuration
Axon will configure a SimpleQueryBus if no QueryBus implementation is explicitly defined in the Application Context. This QueryBus will use the TransactionManager to manage transactions.
Transaction Manager Configuration
If no TransactionManager implementation is explicitly defined in the Application Content, Axon will look for the Spring PlatformTransactionManager bean and wrap that in a TransactionManager. If the Spring bean is not available, the NoOpTransactionManager will be used.
Serializer Configuration
By default, Axon uses an XStream based serializer to serialize objects, as is described in further detail in the Advanced Customizations section. This can be changed by defining a bean of type Serializer in the application context.
While the default Serializer provides an arguably ugly xml based format, it is capable of serializing and deserializing virtually anything, making it a very sensible default. However, for events, which needs to be stored for a long time and perhaps shared across application boundaries, it is desirable to customize the format.
You can define a separate Serializer to be used to serialize events, by assigning it the eventSerializer qualifier. Axon will consider a bean with this qualifier to be the event serializer. If no other bean is defined, Axon will use the default serializer for all other objects to serialize.
Example:
1
class SerializerConfiguration {
2
​
3
    @Qualifier("eventSerializer")
4
    @Bean
5
    public Serializer eventSerializer() {
6
        return new JacksonSerializer();
7
    }
8
}
Copied!
Equal to events, you can also customize the Message Serializer used by your application. The Message Serializer comes into play when your Command and Query message are sent from one node to another in a distributed environment. To set a custom Serializer for you message you can simply define a messageSerializer bean like so:
1
class SerializerConfiguration {
2
​
3
    @Qualifier("messageSerializer")
4
    @Bean
5
    public Serializer messageSerializer() {
6
        return new XStreamSerializer();
7
    }
8
}
Copied!
When overriding both the default serializer and defining an event serializer, we must instruct Spring that the default serializer is, well, the default:
1
class SerializerConfiguration {
2
​
3
    @Primary // Marking it primary means this is the one to use, if no specific Qualifier is requested
4
    @Bean
5
    public Serializer serializer() {
6
        return new MyCustomSerializer();
7
    }
8
​
9
    @Qualifier("eventSerializer")
10
    @Bean
11
    public Serializer eventSerializer() {
12
        return new JacksonSerializer();
13
    }
14
}
Copied!
Aggregate Configuration
The @Aggregate annotation (in package org.axonframework.spring.stereotype) triggers AutoConfiguration to set up the necessary components to use the annotated type as an Aggregate. Note that only the Aggregate Root needs to be annotated.
Axon will automatically register all the @CommandHandler annotated methods with the Command Bus and set up a repository if none is present.
It is possible to define a custom SnapshotTriggerDefinition for an aggregate as a spring bean. In order to tie the SnapshotTriggerDefinition bean to an aggregate, use the snapshotTriggerDefinition attribute on @Aggregate annotation. Listing below shows how to define a custom EventCountSnapshotTriggerDefinition which will take a snapshot on each 500th event.
Note that a Snapshotter instance, if not explicitly defined as a Bean already, will be automatically configured for you. This means you can simply pass the Snapshotter as a parameter to your SnapshotTriggerDefinition.
1
@Bean
2
public SnapshotTriggerDefinition mySnapshotTriggerDefinition(Snapshotter snapshotter) {
3
    return new EventCountSnapshotTriggerDefinition(snapshotter, 500);
4
}
5
...
6
@Aggregate(snapshotTriggerDefinition = "mySnapshotTriggerDefinition")
7
public class MyAggregate {...}
Copied!
Defining a CommandTargetResolver as a bean in the Spring Application context will cause that resolver to be used for all aggregate definitions. However, you can also define multiple beans and specify the instance to use with the commandTargetResolver attribute on @Aggregate annotation will override this behavior. You can for example define a MetaDataCommandTargetResolver which will look for myAggregateId key in meta-data is listed below together with assignment to the aggregate.
1
@Bean
2
public CommandTargetResolver myCommandTargetResolver() {
3
    return new MetaDataCommandTargetResolver("myAggregateId");
4
}
5
...
6
@Aggregate(commandTargetResolver = "myCommandTargetResolver")
7
public class MyAggregate {...}
Copied!
To fully customize the repository used, you can define one in the Application Context. For Axon Framework to use this repository for the intended Aggregate, define the bean name of the repository in the repository attribute on @Aggregate Annotation. Alternatively, specify the bean name of the Repository to be the aggregate's name, (first character lowercase), suffixed with Repository. So on a class of type MyAggregate, the default Repository name is myAggregateRepository. If no bean with that name is found, Axon will define an EventSourcingRepository (which fails if no EventStore is available).
1
@Bean
2
public Repository<MyAggregate> repositoryForMyAggregate(EventStore eventStore) {
3
    return new EventSourcingRepository<>(MyAggregate.class, eventStore);
4
}
5
...
6
@Aggregate(repository = "repositoryForMyAggregate")
7
public class MyAggregate {...}
Copied!
Note that this requires full configuration of the Repository, including any SnapshotTriggerDefinition or AggregateFactory that may otherwise have been configured automatically.
Saga Configuration
The configuration of infrastructure components to operate Sagas is triggered by the @Saga annotation (in package org.axonframework.spring.stereotype). Axon will configure a SagaManager and SagaRepository. The SagaRepository will use a SagaStore available in the context (defaulting to JPASagaStore if JPA is found) for the actual storage of Sagas.
To use different SagaStores for Sagas, provide the bean name of the SagaStore to use in the sagaStore attribute of each @Saga annotation.
Sagas will have resources injected from the application context. Note that this doesn't mean Spring-injecting is used to inject these resources. The @Autowired and @javax.inject.Inject annotation can be used to demarcate dependencies, but they are injected by Axon by looking for these annotations on Fields and Methods. Constructor injection is not (yet) supported.
To tune the configuration of Sagas, it is possible to define a custom SagaConfiguration bean. For an annotated Saga class, Axon will attempt to find a configuration for that Saga. It does so by checking for a bean of type SagaConfiguration with a specific name. For a Saga class called MySaga, the bean that Axon looks for is mySagaConfiguration. If no such bean is found, it creates a Configuration based on available components.
If a SagaConfiguration instance is present for an annotated Saga, that configuration is used to retrieve and register the components for this type of Saga. If the SagaConfiguration bean is not named as described above, it is possible that the Saga is registered twice, and receives events in duplicate. To prevent this, you can specify the bean name of the SagaConfiguration using the @Saga annotation:
1
@Saga(configurationBean = "mySagaConfigBean")
2
public class MySaga {
3
    // methods here 
4
}
5
​
6
// in the Spring configuration:
7
@Bean 
8
public SagaConfiguration<MySaga> mySagaConfigBean() {
9
    // create and return SagaConfiguration instance
10
}
Copied!
Event Handling Configuration
By default, all singleton Spring beans components containing @EventHandler annotated methods will be subscribed to an Event Processor to receive Event Messages published to the Event Bus.
The EventHandlingConfiguration bean, available in the Application Context, has methods to tweak the configuration of the Event Handlers. See Configuration API for details on configuring Event Handlers and Event Processors.
To update the Event Handling Configuration, create an autowired method that set the configuration you desire:
1
@Autowired
2
public void configure(EventHandlingConfiguration config) {
3
    config.usingTrackingProcessors(); // default all processors to tracking mode.
4
}
Copied!
Certain aspect of Event Processors can also be configured in application.properties.
1
axon.eventhandling.processors.name.mode=tracking
2
axon.eventhandling.processors.name.source=eventBus
Copied!
If the name of a processor contains periods ., use the map notation:
1
axon.eventhandling.processors[name].mode=tracking
2
axon.eventhandling.processors[name].source=eventBus
Copied!
Or using application.yml:
1
axon:
2
    eventhandling:
3
        processors:
4
            name:
5
                mode: tracking
6
                source: eventBus
Copied!
The source attribute refers to the name of a bean implementing SubscribableMessageSource or StreamableMessageSource that should be used as the source of events for the mentioned processor. The source default to the Event Bus or Event Store defined in the application context.
Query Handling Configuration
All singleton Spring beans are scanned for methods that have the @QueryHandler annotation. For each method that is found, a new query handler is registered with the query bus.
Parallel processing
Tracking Processors can use multiple threads to process events in parallel. Not all threads need to run on the same node.
One can configure the number of threads (on this instance) as well as the initial number of segments that a processor should define, if non are yet available.
1
axon.eventhandling.processors.name.mode=tracking
2
# Sets the number of maximum number threads to start on this node
3
axon.eventhandling.processors.name.threadCount=2
4
# Sets the initial number of segments (i.e. defines the maximum number of overall threads)
5
axon.eventhandling.processors.name.initialSegmentCount=4
Copied!
Enabling AMQP
To enable AMQP support, ensure that the axon-amqp module is on the classpath and an AMQP ConnectionFactory is available in the application context (e.g. by including the spring-boot-starter-amqp).
To forward Events generated in the application to an AMQP Channel, a single line of application.properties configuration is sufficient:
1
axon.amqp.exchange=ExchangeName
Copied!
This will automatically send all published events to the AMQP Exchange with the given name.
By default, no AMQP transactions are used when sending. This can be overridden using the axon.amqp.transaction-mode property, and setting it to transactional or publisher-ack.
To receive events from a queue and process them inside an Axon application, you need to configure a SpringAMQPMessageSource:
1
@Bean
2
public SpringAMQPMessageSource myQueueMessageSource(AMQPMessageConverter messageConverter) {
3
    return new SpringAMQPMessageSource(messageConverter) {
4
​
5
        @RabbitListener(queues = "myQueue")
6
        @Override
7
        public void onMessage(Message message, Channel channel) throws Exception {
8
            super.onMessage(message, channel);
9
        }
10
    };
11
}
Copied!
and then configure a processor to use this bean as the source for its messages:
1
axon.eventhandling.processors.name.source=myQueueMessageSource
Copied!
Distributing commands
Configuring a distributed command bus can (mostly) be done without any modifications in Configuration files.
First of all, the starters for one of the Axon Distributed Command Bus modules needs to be included (e.g. JGroups or SpringCloud).
Once that is present, a single property needs to be added to the application context, to enable the distributed command bus:
1
axon.distributed.enabled=true
Copied!
There in one setting that is independent of the type of connector used:
1
axon.distributed.load-factor=100
Copied!
Axon will automatically configure a DistributedCommandBus when a CommandRouter as well as a CommandBusConnector are present in the application context. In such case, specifying axon.distributed.enabled isn't even necessary. The latter merely enables autoconfiguration of these routers and connectors.
Using JGroups
This module uses JGroups to detect and communicate with other nodes. The AutoConfiguration will set up the JGroupsConnector using default settings, that may need to be adapted to suit your environment.
By default, the JGroupsConnector will attempt to locate a GossipRouter on the localhost, port 12001.
The settings for the JGroups connector are all prefixed with axon.distributed.jgroups.
1
# the address to bind this instance to. By default, attempts to find the Global IP address
2
axon.distributed.jgroups.bind-addr=GLOBAL
3
# the port to bind the local instance to
4
axon.distributed.jgroups.bind-port=7800
5
​
6
# the name of the JGroups Cluster to connect to
7
axon.distributed.jgroups.cluster-name=Axon
8
​
9
# the JGroups Configuration file to configure JGroups with
10
axon.distributed.jgroups.configuration-file=default_tcp_gossip.xml
11
​
12
# The IP and port of the Gossip Servers (comma separated) to connect to
13
axon.distributed.jgroups.gossip.hosts=localhost[12001]
14
# when true, will start an embedded Gossip Server on bound to the port of the first mentioned gossip host.
15
axon.distributed.jgroups.gossip.auto-start=false
Copied!
The JGroups Configuration file can be use for much more fine-grained control of the Connector's behavior. Check out JGroups's Reference Guide for more information.
Using Spring Cloud
Spring Cloud comes with nice abstractions on top of discovery. Axon can use these abstractions to report its availability and find other Command Bus nodes. For communication with these nodes, Axon uses Spring HTTP, by default.
The Spring Cloud AutoConfiguration doesn't have much to configure. It uses an existing Spring Cloud Discovery Client (so make sure @EnableDiscoveryClient is used and the necessary client is on the classpath).
However, some discovery clients aren't able to update instance metadata dynamically on the server. If Axon detects this, it will automatically fall back to querying that node using HTTP. This is done once on each discovery heartbeat (usually 30 seconds).
This behavior can be configured or disabled, using the following settings in appplication.properties:
1
# whether to fall back to http when no meta-data is available
2
axon.distributed.spring-cloud.fallback-to-http-get=true
3
# the URL on which to publish local data and retrieve from other nodes.
4
axon.distributed.spring-cloud.fallback-url=/message-routing-information
Copied!
For more fine-grained control, provide a SpringCloudHttpBackupCommandRouter or SpringCloudCommandRouter bean in your application context.
Blacklisting
On each heartbeat the memberships of all the nodes in the cluster are updated. If message routing information of a given service instance is not available on this heartbeat signal, that specific service instance gets blacklisted. That blacklisted service instance will be removed from the blacklist if it is no longer present on thereon following heartbeats.
Note
It is regarded as good practice to assign a random value to every service instance name. In doing so, if a given service instance is restarted, it will receive a different name which will mitigate unnecessary blacklisting of nodes.