Spring Integration Extension for Amazon Web Services (AWS)

Introduction

Amazon Web Services (AWS)

Launched in 2006, Amazon Web Services (AWS) provides key infrastructure services for business through its cloud computing platform. Using cloud computing businesses can adopt a new business model whereby they do not have to plan and invest in procuring their own IT infrastructure. They can use the infrastructure and services provided by the cloud service provider and pay as they use the services. Visit AWS Products for more details about various products offered by Amazon as a part their cloud computing services.

Spring Integration Extension for Amazon Web Services provides Spring Integration adapters for the various services provided by the AWS SDK for Java. Note the Spring Integration AWS Extension is based on the Spring Cloud AWS project.

Spring Integration's extensions to AWS

The current project version is 3.0.x and it requires minimum Java 17 and Spring Integration 6.1.x & Spring Cloud AWS 3.0.x. This version is also fully based on AWS Java SDK v2. Therefore, it has a lot of breaking changes since the previous version, for example an XML configuration support was fully removed.

This guide intends to explain briefly the various adapters available for Amazon Web Services such as:

Amazon Simple Storage Service (S3)
Amazon Simple Queue Service (SQS)
Amazon Simple Notification Service (SNS)
Amazon DynamoDB
Amazon Kinesis

Contributing

Pull requests are welcome. Please see the contributor guidelines for details. Additionally, if you are contributing, we recommend following the process for Spring Integration as outlined in the administrator guidelines.

Dependency Management

These dependencies are optional in the project:

io.awspring.cloud:spring-cloud-aws-sns - for SNS channel adapters
io.awspring.cloud:spring-cloud-aws-sqs - for SQS channel adapters
io.awspring.cloud:spring-cloud-aws-s3 - for S3 channel adapters
org.springframework.integration:spring-integration-file - for S3 channel adapters
org.springframework.integration:spring-integration-http - for SNS inbound channel adapter
software.amazon.awssdk:kinesis - for Kinesis channel adapters
software.amazon.kinesis:amazon-kinesis-client - for KCL-based inbound channel adapter
com.amazonaws:amazon-kinesis-producer - for KPL-based MessageHandler
software.amazon.awssdk:dynamodb - for DynamoDbMetadataStore and DynamoDbLockRegistry
software.amazon.awssdk:s3-transfer-manager - for S3MessageHandler
software.amazon.awssdk:aws-crt-client - for S3MessageHandler

Consider to include an appropriate dependency into your project when you use particular component from this project.

Adapters

Amazon Simple Storage Service (Amazon S3)

Introduction

The S3 Channel Adapters are based on the AmazonS3 template and TransferManager. See their specification and JavaDocs for more information.

Inbound Channel Adapter

The S3 Inbound Channel Adapter is represented by the S3InboundFileSynchronizingMessageSource and allows pulling S3 objects as files from the S3 bucket to the local directory for synchronization. This adapter is fully similar to the Inbound Channel Adapters in the FTP and SFTP Spring Integration modules. See more information in the FTP/FTPS Adapters Chapter for common options or SessionFactory, RemoteFileTemplate and FileListFilter abstractions.

The Java Configuration is:

@SpringBootApplication
public static class MyConfiguration {

    @Autowired
    private S3Client amazonS3;

    @Bean
    public S3InboundFileSynchronizer s3InboundFileSynchronizer() {
    	S3InboundFileSynchronizer synchronizer = new S3InboundFileSynchronizer(amazonS3());
    	synchronizer.setDeleteRemoteFiles(true);
    	synchronizer.setPreserveTimestamp(true);
    	synchronizer.setRemoteDirectory(S3_BUCKET);
    	synchronizer.setFilter(new S3RegexPatternFileListFilter(".*\\.test$"));
    	Expression expression = PARSER.parseExpression("#this.toUpperCase() + '.a'");
    	synchronizer.setLocalFilenameGeneratorExpression(expression);
    	return synchronizer;
    }

    @Bean
    @InboundChannelAdapter(value = "s3FilesChannel", poller = @Poller(fixedDelay = "100"))
    public S3InboundFileSynchronizingMessageSource s3InboundFileSynchronizingMessageSource() {
    	S3InboundFileSynchronizingMessageSource messageSource =
    			new S3InboundFileSynchronizingMessageSource(s3InboundFileSynchronizer());
    	messageSource.setAutoCreateLocalDirectory(true);
    	messageSource.setLocalDirectory(LOCAL_FOLDER);
    	messageSource.setLocalFilter(new AcceptOnceFileListFilter<File>());
    	return messageSource;
    }

    @Bean
    public PollableChannel s3FilesChannel() {
    	return new QueueChannel();
    }
}

With this config you receive messages with java.io.File payload from the s3FilesChannel after periodic synchronization of content from the Amazon S3 bucket into the local directory.

Streaming Inbound Channel Adapter

This adapter produces message with payloads of type InputStream, allowing S3 objects to be fetched without writing to the local file system. Since the session remains open, the consuming application is responsible for closing the session when the file has been consumed. The session is provided in the closeableResource header (IntegrationMessageHeaderAccessor.CLOSEABLE_RESOURCE). Standard framework components, such as the FileSplitter and StreamTransformer will automatically close the session.

The following Spring Boot application provides an example of configuring the S3 inbound streaming adapter using Java configuration:

@SpringBootApplication
public class S3JavaApplication {

    public static void main(String[] args) {
        new SpringApplicationBuilder(S3JavaApplication.class)
            .web(false)
            .run(args);
    }
    
    @Autowired
    private S3Client amazonS3;

    @Bean
    @InboundChannelAdapter(value = "s3Channel", poller = @Poller(fixedDelay = "100"))
    public MessageSource<InputStream> s3InboundStreamingMessageSource() {    
        S3StreamingMessageSource messageSource = new S3StreamingMessageSource(template());
        messageSource.setRemoteDirectory(S3_BUCKET);
        messageSource.setFilter(new S3PersistentAcceptOnceFileListFilter(new SimpleMetadataStore(),
                                   "streaming"));    	
    	return messageSource;
    }

    @Bean
    @Transformer(inputChannel = "s3Channel", outputChannel = "data")
    public org.springframework.integration.transformer.Transformer transformer() {
        return new StreamTransformer();
    }
    
    @Bean
    public S3RemoteFileTemplate template() {
        return new S3RemoteFileTemplate(new S3SessionFactory(amazonS3));
    }

    @Bean
    public PollableChannel s3Channel() {
    	return new QueueChannel();
    }
}

NOTE: Unlike the non-streaming inbound channel adapter, this adapter does not prevent duplicates by default. If you do not delete the remote file and wish to prevent the file being processed again, you can configure an S3PersistentFileListFilter in the filter attribute. If you don’t actually want to persist the state, an in-memory SimpleMetadataStore can be used with the filter. If you wish to use a filename pattern (or regex) as well, use a CompositeFileListFilter.

Outbound Channel Adapter

The S3 Outbound Channel Adapter is represented by the S3MessageHandler and allows performing upload, download and copy (see S3MessageHandler.Command enum) operations in the provided S3 bucket.

The Java Configuration is:

@SpringBootApplication
public static class MyConfiguration {

    @Autowired
    private S3AsyncClient amazonS3;

    @Bean
    @ServiceActivator(inputChannel = "s3UploadChannel")
    public MessageHandler s3MessageHandler() {
    	return new S3MessageHandler(amazonS3(), "my-bucket");
    }

}

With this config you can send a message with the java.io.File as payload and the transferManager.upload() operation will be performed, where the file name is used as a S3 Object key.

See more information in the S3MessageHandler JavaDocs.

NOTE: The AWS SDK recommends to use S3CrtAsyncClient for S3TransferManager, therefore an S3AsyncClient.crtBuilder() has to be used to achieve respective upload and download requirements.

Outbound Gateway

The S3 Outbound Gateway is represented by the same S3MessageHandler with the produceReply = true constructor argument for Java Configuration.

The "request-reply" nature of this gateway is async and the Transfer result from the TransferManager operation is sent to the outputChannel, assuming the transfer progress observation in the downstream flow.

The TransferListener can be supplied via AwsHeaders.TRANSFER_LISTENER header of the request message to track the transfer progress.

See more information in the S3MessageHandler JavaDocs.

Simple Email Service (SES)

There is no adapter for SES, since Spring Cloud AWS provides implementations for org.springframework.mail.MailSender - SimpleEmailServiceMailSender and SimpleEmailServiceJavaMailSender, which can be injected to the MailSendingMessageHandler.

Amazon Simple Queue Service (SQS)

The SQS adapters are fully based on the Spring Cloud AWS foundation, so for more information about the background components and core configuration, please, refer to the documentation of that project.

Outbound Channel Adapter

The SQS Outbound Channel Adapter is presented by the SqsMessageHandler implementation and allows sending messages to the SQS queue with provided SqsAsyncClient client. An SQS queue can be configured explicitly on the adapter (using org.springframework.integration.expression.ValueExpression) or as a SpEL Expression, which is evaluated against request message as a root object of evaluation context. In addition, the queue can be extracted from the message headers under AwsHeaders.QUEUE.

The Java Configuration is pretty simple:

@SpringBootApplication
public static class MyConfiguration {

    @Bean
    @ServiceActivator(inputChannel = "sqsSendChannel")
    public MessageHandler sqsMessageHandler(SqsAsyncClient amazonSqs) {
    	return new SqsMessageHandler(amazonSqs);
    }

}

Starting with version 2.0, the SqsMessageHandler can be configured with the HeaderMapper to map message headers to the SQS message attributes. See SqsHeaderMapper implementation for more information and also consult with Amazon SQS Message Attributes about value types and restrictions.

Inbound Channel Adapter

The SQS Inbound Channel Adapter is a message-driven implementation for the MessageProducer and is represented with SqsMessageDrivenChannelAdapter. This channel adapter is based on the io.awspring.cloud.sqs.listener.SqsMessageListenerContainer to receive messages from the provided queues in async manner and send an enhanced Spring Integration Message to the provided MessageChannel.

The Java Configuration is pretty simple:

@SpringBootApplication
public static class MyConfiguration {

	@Autowired
	private SqsAsyncClient amazonSqs;

	@Bean
	public PollableChannel inputChannel() {
		return new QueueChannel();
	}

	@Bean
	public MessageProducer sqsMessageDrivenChannelAdapter() {
		SqsMessageDrivenChannelAdapter adapter = new SqsMessageDrivenChannelAdapter(this.amazonSqs, "myQueue");
		adapter.setOutputChannel(inputChannel());
		return adapter;
	}
}

The target listener container can be configured via SqsMessageDrivenChannelAdapter.setSqsContainerOptions(SqsContainerOptions) option.

Amazon Simple Notification Service (SNS)

Amazon SNS is a publish-subscribe messaging system that allows clients to publish notification to a particular topic. Other interested clients may subscribe using different protocols like HTTP/HTTPS, e-mail, or an Amazon SQS queue to receive the messages. Plus mobile devices can be registered as subscribers from the AWS Management Console.

Unfortunately Spring Cloud AWS doesn't provide flexible components which can be used from the channel adapter implementations, but Amazon SNS API is pretty simple, on the other hand. Hence, Spring Integration AWS SNS Support is straightforward and just allows to provide channel adapter foundation for Spring Integration applications.

Since e-mail, SMS and mobile device subscription/unsubscription confirmation is out of the Spring Integration application scope and can be done only from the AWS Management Console, we provide only HTTP/HTTPS SNS endpoint in face of SnsInboundChannelAdapter. The SQS-to-SNS subscription can also be done through account configuration: https://docs.aws.amazon.com/sns/latest/dg/subscribe-sqs-queue-to-sns-topic.html.

Inbound Channel Adapter

The SnsInboundChannelAdapter is an extension of HttpRequestHandlingMessagingGateway and must be as a part of Spring MVC application. Its URL must be used from the AWS Management Console to add this endpoint as a subscriber to the SNS Topic. However. before receiving any notification itself this HTTP endpoint must confirm the subscription.

See SnsInboundChannelAdapter JavaDocs for more information.

An important option of this adapter to consider is handleNotificationStatus. This boolean flag indicates if the adapter should send SubscriptionConfirmation/UnsubscribeConfirmation message to the output-channel or not. If that the AwsHeaders.NOTIFICATION_STATUS message header is present in the message with the NotificationStatus object, which can be used in the downstream flow to confirm subscription or not. Or "re-confirm" it in case of UnsubscribeConfirmation message.

In addition, the AwsHeaders#SNS_MESSAGE_TYPE message header is represented to simplify a routing in the downstream flow.

The Java Configuration is pretty simple:

@SpringBootApplication
public static class MyConfiguration {

	@Autowired
	private SnsClient amazonSns;

	@Bean
    public PollableChannel inputChannel() {
    	return new QueueChannel();
    }

    @Bean
    public HttpRequestHandler sqsMessageDrivenChannelAdapter() {
    	SnsInboundChannelAdapter adapter = new SnsInboundChannelAdapter(amazonSns(), "/mySampleTopic");
    	adapter.setRequestChannel(inputChannel());
    	adapter.setHandleNotificationStatus(true);
    	return adapter;
    }
}

Note: by default the message payload is a Map converted from the received Topic JSON message. For the convenience a payload-expression is provided with the Message as a root object of the evaluation context. Hence, even some HTTP headers, populated by the DefaultHttpHeaderMapper, are available for the evaluation context.

Outbound Channel Adapter

The SnsMessageHandler is a simple one-way Outbound Channel Adapter to send Topic Notification using SnsAsyncClient service.

This Channel Adapter (MessageHandler) accepts these options:

topic-arn (topic-arn-expression) - the SNS Topic to send notification for.
subject (subject-expression) - the SNS Notification Subject;
body-expression - the SpEL expression to evaluate the message property for the software.amazon.awssdk.services.sns.model.PublishRequest.
resource-id-resolver - a ResourceIdResolver bean reference to resolve logical topic names to physical resource ids;

See SnsMessageHandler JavaDocs for more information.

The Java Config looks like:

@Bean
public MessageHandler snsMessageHandler() {
    SnsMessageHandler handler = new SnsMessageHandler(amazonSns());
    handler.setTopicArn("arn:aws:sns:eu-west:123456789012:test");
    String bodyExpression = "T(SnsBodyBuilder).withDefault(payload).forProtocols(payload.substring(0, 140), 'sms')";
    handler.setBodyExpression(spelExpressionParser.parseExpression(bodyExpression));

    // message-group ID and deduplication ID are used for FIFO topics
    handler.setMessageGroupId("foo-messages");
    String deduplicationExpression = "headers.id";
    handler.setMessageDeduplicationIdExpression(spelExpressionParser.parseExpression(deduplicationExpression))''
    return handler;
}

NOTE: the bodyExpression can be evaluated to a org.springframework.integration.aws.support.SnsBodyBuilder allowing the configuration of a json messageStructure for the PublishRequest and provide separate messages for different protocols. The same SnsBodyBuilder rule is applied for the raw payload if the bodyExpression hasn't been configured. NOTE: if the payload of requestMessage is a software.amazon.awssdk.services.sns.model.PublishRequest already, the SnsMessageHandler doesn't do anything with it, and it is sent as-is.

Starting with version 2.0, the SnsMessageHandler can be configured with the HeaderMapper to map message headers to the SNS message attributes. See SnsHeaderMapper implementation for more information and also consult with Amazon SNS Message Attributes about value types and restrictions.

Starting with version 2.5.3, the SnsMessageHandler supports sending to SNS FIFO topics using the messageGroupId/messageGroupIdExpression and messageDeduplicationIdExpression properties.

Metadata Store for Amazon DynamoDB

The DynamoDbMetadataStore, a ConcurrentMetadataStore implementation, is provided to keep the metadata for Spring Integration components in the distributed Amazon DynamoDB store. The implementation is based on a simple table with metadataKey and metadataValue attributes, both are string types and the metadataKey is partition key of the table. By default, the SpringIntegrationMetadataStore table is used, and it is created during DynamoDbMetaDataStore initialization if that doesn't exist yet. The DynamoDbMetadataStore can be used for the KinesisMessageDrivenChannelAdapter as a cloud-based cehckpointStore.

Starting with version 2.0, the DynamoDbMetadataStore can be configured with the timeToLive option to enable the DynamoDB TTL feature. The expireAt attribute is added to each item with the value based on the sum of current time and provided timeToLive in seconds. If the provided timeToLive value is non-positive, the TTL functionality is disabled on the table.

Amazon Kinesis

Amazon Kinesis is a platform for streaming data on AWS, making it easy to load and analyze streaming data, and also providing the ability for you to build custom streaming data applications for specialized needs.

Inbound Channel Adapter

The KinesisMessageDrivenChannelAdapter is an extension of the MessageProducerSupport - event-driver channel adapter.

See KinesisMessageDrivenChannelAdapter JavaDocs and its setters for more information how to use and how to configure it in the application for Kinesis streams ingestion.

The Java Configuration is pretty simple:

@SpringBootApplication
public static class MyConfiguration {

    @Bean
    public KinesisMessageDrivenChannelAdapter kinesisInboundChannelChannel(KinesisAsyncClient amazonKinesis) {
        KinesisMessageDrivenChannelAdapter adapter =
            new KinesisMessageDrivenChannelAdapter(amazonKinesis, "MY_STREAM");
        adapter.setOutputChannel(kinesisReceiveChannel());
        return adapter;
    }
}

This channel adapter can be configured with the DynamoDbMetadataStore mentioned above to track sequence checkpoints for shards in the cloud environment when we have several instances of our Kinesis application. By default, this adapter uses DeserializingConverter to convert byte[] from the Record data. Can be specified as null with meaning no conversion and the target Message is sent with the byte[] payload.

Additional headers like AwsHeaders.RECEIVED_STREAM, AwsHeaders.SHARD, AwsHeaders.RECEIVED_PARTITION_KEY and AwsHeaders.RECEIVED_SEQUENCE_NUMBER are populated to the message for downstream logic. When CheckpointMode.manual is used the Checkpointer instance is populated to the AwsHeaders.CHECKPOINTER header for an acknowledgment in the downstream logic manually.

The KinesisMessageDrivenChannelAdapter can be configured with the ListenerMode record or batch to process records one by one or send the whole just polled batch of records. If Converter is configured to null, the entire List<Record> is sent as a payload. Otherwise, a list of converted Record.getData().array() is wrapped to the payload of message to send. In this case the AwsHeaders.RECEIVED_PARTITION_KEY and AwsHeaders.RECEIVED_SEQUENCE_NUMBER headers contains values as a List<String> of partition keys and sequence numbers of converted records respectively.

The consumer group is included to the metadata store key. When records are consumed, they are filtered by the last stored lastCheckpoint under the key as [CONSUMER_GROUP]:[STREAM]:[SHARD_ID].

Starting with version 2.0, the KinesisMessageDrivenChannelAdapter can be configured with the InboundMessageMapper to extract message headers embedded into the record data (if any). See EmbeddedJsonHeadersMessageMapper implementation for more information. When InboundMessageMapper is used together with the ListenerMode.batch, each Record is converted to the Message with extracted embedded headers (if any) and converted byte[] payload if any and converter is present. In this case AwsHeaders.RECEIVED_PARTITION_KEY and AwsHeaders.RECEIVED_SEQUENCE_NUMBER headers are populated to the particular message for a record. These messages are wrapped as a list payload to one outbound message.

Starting with version 2.0, the KinesisMessageDrivenChannelAdapter can be configured with the LockRegistry for leader selection for the provided shards or derived from the provided streams. The KinesisMessageDrivenChannelAdapter iterates over its shards and tries to acquire a distributed lock for the shard in its consumer group. If LockRegistry is not provided, no exclusive locking happens and all the shards are consumed by this KinesisMessageDrivenChannelAdapter. See also DynamoDbLockRegistry for more information.

The KinesisMessageDrivenChannelAdapter can be configured with a Function<List<Shard>, List<Shard>> shardListFilter to filter the available, open, non-exhausted shards. This filter Function will be called each time the shard list is refreshed.

For example, users may want to fully read any parent shards before starting to read their child shards. This could be achieved as follows:

    openShards -> {
        Set<String> openShardIds = openShards.stream().map(Shard::getShardId).collect(Collectors.toSet());
        // only return open shards which have no parent available for reading
        return openShards.stream()
                .filter(shard -> !openShardIds.contains(shard.getParentShardId())
                        && !openShardIds.contains(shard.getAdjacentParentShardId()))
                .collect(Collectors.toList());
        }

Starting with version 3.0, any exception thrown from the record process may lead to shard iterator rewinding to the latest check-pointed sequence or the first one in the current failed batch. This ensures an at-least-once delivery for possibly failed records. If the latest checkpoint is equal to the highest sequence in the batch, then shard consumer continue with the next iterator.

Also, the KclMessageDrivenChannelAdapter is provided for performing streams consumption by Kinesis Client Library. See its JavaDocs for more information.

Outbound Channel Adapter

The KinesisMessageHandler is an AbstractMessageHandler to perform put record to the Kinesis stream. The stream, partition key (or explicit hash key) and sequence number can be determined against request message via evaluation provided expressions or can be specified statically. They also can be specified as AwsHeaders.STREAM, AwsHeaders.PARTITION_KEY and AwsHeaders.SEQUENCE_NUMBER respectively.

The KinesisMessageHandler can be configured with the outputChannel for sending a Message on successful put operation. The payload is the original request and additional AwsHeaders.SHARD and AwsHeaders.SEQUENCE_NUMBER headers are populated from the PutRecordResult. If the request payload is a PutRecordsRequest, the full PutRecordsResult is populated in the AwsHeaders.SERVICE_RESULT header instead.

When an async failure is happened on the put operation, the ErrorMessage is sent to the errorChannel header or global one. The payload is an AwsRequestFailureException.

The payload of request message can be:

PutRecordsRequest to perform KinesisAsyncClient.putRecords
PutRecordRequest to perform KinesisAsyncClient.putRecord
ByteBuffer to represent a data of the PutRecordRequest
byte[] which is wrapped to the ByteBuffer
any other type which is converted to the byte[] by the provided Converter; the SerializingConverter is used by default.

The Java Configuration for the message handler:

@Bean
@ServiceActivator(inputChannel = "kinesisSendChannel")
public MessageHandler kinesisMessageHandler(KinesisAsyncClient amazonKinesis,
                                            MessageChannel channel) {
    KinesisMessageHandler kinesisMessageHandler = new KinesisMessageHandler(amazonKinesis);
    kinesisMessageHandler.setPartitionKey("1");
    kinesisMessageHandler.setOutputChannel(channel);
    return kinesisMessageHandler;
}

Starting with version 2.0, the KinesisMessageHandler can be configured with the OutboundMessageMapper to embed message headers into the record data alongside with the payload. See EmbeddedJsonHeadersMessageMapper implementation for more information.

Also, the KplMessageHandler is provided for performing streams consumption by Kinesis Producer Library.

Lock Registry for Amazon DynamoDB

Starting with version 2.0, the DynamoDbLockRegistry implementation is available. Certain components (for example aggregator and resequencer) use a lock obtained from a LockRegistry instance to ensure that only one thread is manipulating a group at a time. The DefaultLockRegistry performs this function within a single component; you can now configure an external lock registry on these components. When used with a shared MessageGroupStore, the DynamoDbLockRegistry can be used to provide this functionality across multiple application instances, such that only one instance can manipulate the group at a time.
This implementation can also be used for the distributed leader elections using a LockRegistryLeaderInitiator.

Testing

The tests in the project are performed via Testcontainers and Local Stack image. See LocalstackContainerTest interface Javadocs for more information.