Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Integrating with Amazon Glue Schema Registry

These sections describe integrations with Amazon Glue Schema Registry. The examples in these section show a schema with AVRO data format. For more examples, including schemas with JSON data format, see the integration tests and ReadMe information in the Amazon Glue Schema Registry open source repository.

Use case: Connecting Schema Registry to Amazon MSK or Apache Kafka

Let's assume you are writing data to an Apache Kafka topic, and you can follow these steps to get started.

Use case: Integrating Amazon Kinesis Data Streams with the Amazon Glue Schema Registry

This integration requires that you have an existing Amazon Kinesis data stream. For more information, see Getting Started with Amazon Kinesis Data Streams in the Amazon Kinesis Data Streams Developer Guide.

There are two ways that you can interact with data in a Kinesis data stream.

If you currently use the KPL/KCL libraries, we recommend continuing to use that method. There are updated KCL and KPL versions with Schema Registry integrated, as shown in the examples. Otherwise, you can use the sample code to leverage the Amazon Glue Schema Registry if using the KDS APIs directly.

Schema Registry integration is only available with KPL v0.14.2 or later and with KCL v2.3 or later. Schema Registry integration with JSON data format is available with KPL v0.14.8 or later and with KCL v2.3.6 or later.

Interacting with Data Using Kinesis SDK V2

This section describes interacting with Kinesis using Kinesis SDK V2

// Example JSON Record, you can construct a AVRO record also
private static final JsonDataWithSchema record = JsonDataWithSchema.builder(schemaString, payloadString);
private static final DataFormat dataFormat = DataFormat.JSON;

//Configurations for Schema Registry
GlueSchemaRegistryConfiguration gsrConfig = new GlueSchemaRegistryConfiguration("us-east-1");

GlueSchemaRegistrySerializer glueSchemaRegistrySerializer =
        new GlueSchemaRegistrySerializerImpl(awsCredentialsProvider, gsrConfig);
GlueSchemaRegistryDataFormatSerializer dataFormatSerializer =
        new GlueSchemaRegistrySerializerFactory().getInstance(dataFormat, gsrConfig);

Schema gsrSchema =
        new Schema(dataFormatSerializer.getSchemaDefinition(record), dataFormat.name(), "MySchema");

byte[] serializedBytes = dataFormatSerializer.serialize(record);

byte[] gsrEncodedBytes = glueSchemaRegistrySerializer.encode(streamName, gsrSchema, serializedBytes);

PutRecordRequest putRecordRequest = PutRecordRequest.builder()
        .streamName(streamName)
        .partitionKey("partitionKey")
        .data(SdkBytes.fromByteArray(gsrEncodedBytes))
        .build();
shardId = kinesisClient.putRecord(putRecordRequest)
        .get()
        .shardId();
        
GlueSchemaRegistryDeserializer glueSchemaRegistryDeserializer = new GlueSchemaRegistryDeserializerImpl(awsCredentialsProvider, gsrConfig);

GlueSchemaRegistryDataFormatDeserializer gsrDataFormatDeserializer =
        glueSchemaRegistryDeserializerFactory.getInstance(dataFormat, gsrConfig);

GetShardIteratorRequest getShardIteratorRequest = GetShardIteratorRequest.builder()
        .streamName(streamName)
        .shardId(shardId)
        .shardIteratorType(ShardIteratorType.TRIM_HORIZON)
        .build();

String shardIterator = kinesisClient.getShardIterator(getShardIteratorRequest)
        .get()
        .shardIterator();

GetRecordsRequest getRecordRequest = GetRecordsRequest.builder()
        .shardIterator(shardIterator)
        .build();
GetRecordsResponse recordsResponse = kinesisClient.getRecords(getRecordRequest)
        .get();

List<Object> consumerRecords = new ArrayList<>();
List<Record> recordsFromKinesis = recordsResponse.records();

for (int i = 0; i < recordsFromKinesis.size(); i++) {
    byte[] consumedBytes = recordsFromKinesis.get(i)
            .data()
            .asByteArray();

    Schema gsrSchema = glueSchemaRegistryDeserializer.getSchema(consumedBytes);
    Object decodedRecord = gsrDataFormatDeserializer.deserialize(ByteBuffer.wrap(consumedBytes),
                                                                    gsrSchema.getSchemaDefinition());
    consumerRecords.add(decodedRecord);
}

Interacting with data using the KPL/KCL libraries

This section describes integrating Kinesis Data Streams with Schema Registry using the KPL/KCL libraries. For more information on using KPL/KCL, see Developing Producers Using the Amazon Kinesis Producer Library in the Amazon Kinesis Data Streams Developer Guide.

Setting up the Schema Registry in KPL

Setting up the Kinesis client library

You will develop your Kinesis Client Library consumer in Java. For more information, see Developing a Kinesis Client Library Consumer in Java in the Amazon Kinesis Data Streams Developer Guide.

Interacting with data using the Kinesis Data Streams APIs

This section describes integrating Kinesis Data Streams with Schema Registry using the Kinesis Data Streams APIs.

Interacting with data using the Kinesis Data Streams APIs

The following is example code for using the PutRecords and GetRecords APIs.

//Full sample code
import com.amazonaws.services.schemaregistry.deserializers.GlueSchemaRegistryDeserializerImpl;
import com.amazonaws.services.schemaregistry.serializers.GlueSchemaRegistrySerializerImpl;
import com.amazonaws.services.schemaregistry.utils.AVROUtils;
import com.amazonaws.services.schemaregistry.utils.AWSSchemaRegistryConstants;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericDatumReader;
import org.apache.avro.generic.GenericDatumWriter;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.io.Decoder;
import org.apache.avro.io.DecoderFactory;
import org.apache.avro.io.Encoder;
import org.apache.avro.io.EncoderFactory;
import software.amazon.awssdk.auth.credentials.DefaultCredentialsProvider;
import software.amazon.awssdk.services.glue.model.DataFormat;

import java.io.ByteArrayOutputStream;
import java.io.File;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.util.Collections;
import java.util.HashMap;
import java.util.Map;


public class PutAndGetExampleWithEncodedData {
    static final String regionName = "us-east-2";
    static final String streamName = "testStream1";
    static final String schemaName = "User-Topic";
    static final String AVRO_USER_SCHEMA_FILE = "src/main/resources/user.avsc";
    KinesisApi kinesisApi = new KinesisApi();

    void runSampleForPutRecord() throws IOException {
        Object testRecord = getTestRecord();
        byte[] recordAsBytes = convertRecordToBytes(testRecord);
        String schemaDefinition = AVROUtils.getInstance().getSchemaDefinition(testRecord);

        //The following lines add a Schema Header to a record
        com.amazonaws.services.schemaregistry.common.Schema awsSchema =
            new com.amazonaws.services.schemaregistry.common.Schema(schemaDefinition, DataFormat.AVRO.name(),
                schemaName);
        GlueSchemaRegistrySerializerImpl glueSchemaRegistrySerializer =
            new GlueSchemaRegistrySerializerImpl(DefaultCredentialsProvider.builder().build(), new GlueSchemaRegistryConfiguration(regionName));
        byte[] recordWithSchemaHeader =
            glueSchemaRegistrySerializer.encode(streamName, awsSchema, recordAsBytes);

        //Use PutRecords api to pass a list of records
        kinesisApi.putRecords(Collections.singletonList(recordWithSchemaHeader), streamName, regionName);

        //OR
        //Use PutRecord api to pass single record
        //kinesisApi.putRecord(recordWithSchemaHeader, streamName, regionName);
    }

    byte[] runSampleForGetRecord() throws IOException {
        ByteBuffer recordWithSchemaHeader = kinesisApi.getRecords(streamName, regionName);

        //The following lines remove the schema registry header
        GlueSchemaRegistryDeserializerImpl glueSchemaRegistryDeserializer =
            new GlueSchemaRegistryDeserializerImpl(DefaultCredentialsProvider.builder().build(), new GlueSchemaRegistryConfiguration(regionName));
        byte[] recordWithSchemaHeaderBytes = new byte[recordWithSchemaHeader.remaining()];
        recordWithSchemaHeader.get(recordWithSchemaHeaderBytes, 0, recordWithSchemaHeaderBytes.length);
        
        com.amazonaws.services.schemaregistry.common.Schema awsSchema =
            glueSchemaRegistryDeserializer.getSchema(recordWithSchemaHeaderBytes);
        
        byte[] record = glueSchemaRegistryDeserializer.getData(recordWithSchemaHeaderBytes);

        //The following lines serialize an AVRO schema record
        if (DataFormat.AVRO.name().equals(awsSchema.getDataFormat())) {
            Schema avroSchema = new org.apache.avro.Schema.Parser().parse(awsSchema.getSchemaDefinition());
            Object genericRecord = convertBytesToRecord(avroSchema, record);
            System.out.println(genericRecord);
        }

        return record;
    }

    private byte[] convertRecordToBytes(final Object record) throws IOException {
        ByteArrayOutputStream recordAsBytes = new ByteArrayOutputStream();
        Encoder encoder = EncoderFactory.get().directBinaryEncoder(recordAsBytes, null);
        GenericDatumWriter datumWriter = new GenericDatumWriter<>(AVROUtils.getInstance().getSchema(record));
        datumWriter.write(record, encoder);
        encoder.flush();
        return recordAsBytes.toByteArray();
    }

    private GenericRecord convertBytesToRecord(Schema avroSchema, byte[] record) throws IOException {
        final GenericDatumReader<GenericRecord> datumReader = new GenericDatumReader<>(avroSchema);
        Decoder decoder = DecoderFactory.get().binaryDecoder(record, null);
        GenericRecord genericRecord = datumReader.read(null, decoder);
        return genericRecord;
    }

    private Map<String, String> getMetadata() {
        Map<String, String> metadata = new HashMap<>();
        metadata.put("event-source-1", "topic1");
        metadata.put("event-source-2", "topic2");
        metadata.put("event-source-3", "topic3");
        metadata.put("event-source-4", "topic4");
        metadata.put("event-source-5", "topic5");
        return metadata;
    }

    private GlueSchemaRegistryConfiguration getConfigs() {
        GlueSchemaRegistryConfiguration configs = new GlueSchemaRegistryConfiguration(regionName);
        configs.setSchemaName(schemaName);
        configs.setAutoRegistration(true);
        configs.setMetadata(getMetadata());
        return configs;
    }

    private Object getTestRecord() throws IOException {
        GenericRecord genericRecord;
        Schema.Parser parser = new Schema.Parser();
        Schema avroSchema = parser.parse(new File(AVRO_USER_SCHEMA_FILE));

        genericRecord = new GenericData.Record(avroSchema);
        genericRecord.put("name", "testName");
        genericRecord.put("favorite_number", 99);
        genericRecord.put("favorite_color", "red");

        return genericRecord;
    }
}

Use case: Amazon Managed Service for Apache Flink

Apache Flink is a popular open source framework and distributed processing engine for stateful computations over unbounded and bounded data streams. Amazon Managed Service for Apache Flink is a fully managed Amazon service that enables you to build and manage Apache Flink applications to process streaming data.

Open source Apache Flink provides a number of sources and sinks. For example, predefined data sources include reading from files, directories, and sockets, and ingesting data from collections and iterators. Apache Flink DataStream Connectors provide code for Apache Flink to interface with various third-party systems, such as Apache Kafka or Kinesis as sources and/or sinks.

For more information, see Amazon Kinesis Data Analytics Developer Guide.

Apache Flink Kafka connector

Apache Flink provides an Apache Kafka data stream connector for reading data from and writing data to Kafka topics with exactly-once guarantees. Flink's Kafka consumer, FlinkKafkaConsumer, provides access to read from one or more Kafka topics. Apache Flink’s Kafka Producer, FlinkKafkaProducer, allows writing a stream of records to one or more Kafka topics. For more information, see Apache Kafka Connector.

Apache Flink Kinesis streams Connector

The Kinesis data stream connector provides access to Amazon Kinesis Data Streams. The FlinkKinesisConsumer is an exactly-once parallel streaming data source that subscribes to multiple Kinesis streams within the same Amazon service region, and can transparently handle re-sharding of streams while the job is running. Each subtask of the consumer is responsible for fetching data records from multiple Kinesis shards. The number of shards fetched by each subtask will change as shards are closed and created by Kinesis. The FlinkKinesisProducer uses Kinesis Producer Library (KPL) to put data from an Apache Flink stream into a Kinesis stream. For more information, see Amazon Kinesis Streams Connector.

For more information, see the Amazon Glue Schema Github repository.

Integrating with Apache Flink

The SerDes library provided with Schema Registry integrates with Apache Flink. To work with Apache Flink, you are required to implement SerializationSchema and DeserializationSchema interfaces called GlueSchemaRegistryAvroSerializationSchema and GlueSchemaRegistryAvroDeserializationSchema, which you can plug into Apache Flink connectors.

Adding an Amazon Glue Schema Registry dependency into the Apache Flink application

To set up the integration dependencies to Amazon Glue Schema Registry in the Apache Flink application:

Integrating Kafka or Amazon MSK with Apache Flink

You can use Managed Service for Apache Flink for Apache Flink, with Kafka as a source or Kafka as a sink.

Kafka as a source

The following diagram shows integrating Kinesis Data Streams with Managed Service for Apache Flink for Apache Flink, with Kafka as a source.

Kafka as a sink

The following diagram shows integrating Kinesis Data Streams with Managed Service for Apache Flink for Apache Flink, with Kafka as a sink.

To integrate Kafka (or Amazon MSK) with Managed Service for Apache Flink for Apache Flink, with Kafka as a source or Kafka as a sink, make the code changes below. Add the bolded code blocks to your respective code in the analogous sections.

If Kafka is the source, then use the deserializer code (block 2). If Kafka is the sink, use the serializer code (block 3).

StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

String topic = "topic";
Properties properties = new Properties();
properties.setProperty("bootstrap.servers", "localhost:9092");
properties.setProperty("group.id", "test");

// block 1
Map<String, Object> configs = new HashMap<>();
configs.put(AWSSchemaRegistryConstants.AWS_REGION, "aws-region");
configs.put(AWSSchemaRegistryConstants.SCHEMA_AUTO_REGISTRATION_SETTING, true);
configs.put(AWSSchemaRegistryConstants.AVRO_RECORD_TYPE, AvroRecordType.GENERIC_RECORD.getName());

FlinkKafkaConsumer<GenericRecord> consumer = new FlinkKafkaConsumer<>(
    topic,
    // block 2
    GlueSchemaRegistryAvroDeserializationSchema.forGeneric(schema, configs),
    properties);

FlinkKafkaProducer<GenericRecord> producer = new FlinkKafkaProducer<>(
    topic,
    // block 3
    GlueSchemaRegistryAvroSerializationSchema.forGeneric(schema, topic, configs),
    properties);

DataStream<GenericRecord> stream = env.addSource(consumer);
stream.addSink(producer);
env.execute();

Integrating Kinesis Data Streams with Apache Flink

You can use Managed Service for Apache Flink for Apache Flink with Kinesis Data Streams as a source or a sink.

Kinesis Data Streams as a source

The following diagram shows integrating Kinesis Data Streams with Managed Service for Apache Flink for Apache Flink, with Kinesis Data Streams as a source.

Kinesis Data Streams as a sink

The following diagram shows integrating Kinesis Data Streams with Managed Service for Apache Flink for Apache Flink, with Kinesis Data Streams as a sink.

To integrate Kinesis Data Streams with Managed Service for Apache Flink for Apache Flink, with Kinesis Data Streams as a source or Kinesis Data Streams as a sink, make the code changes below. Add the bolded code blocks to your respective code in the analogous sections.

If Kinesis Data Streams is the source, use the deserializer code (block 2). If Kinesis Data Streams is the sink, use the serializer code (block 3).

StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

String streamName = "stream";
Properties consumerConfig = new Properties();
consumerConfig.put(AWSConfigConstants.AWS_REGION, "aws-region");
consumerConfig.put(AWSConfigConstants.AWS_ACCESS_KEY_ID, "aws_access_key_id");
consumerConfig.put(AWSConfigConstants.AWS_SECRET_ACCESS_KEY, "aws_secret_access_key");
consumerConfig.put(ConsumerConfigConstants.STREAM_INITIAL_POSITION, "LATEST");

// block 1
Map<String, Object> configs = new HashMap<>();
configs.put(AWSSchemaRegistryConstants.AWS_REGION, "aws-region");
configs.put(AWSSchemaRegistryConstants.SCHEMA_AUTO_REGISTRATION_SETTING, true);
configs.put(AWSSchemaRegistryConstants.AVRO_RECORD_TYPE, AvroRecordType.GENERIC_RECORD.getName());

FlinkKinesisConsumer<GenericRecord> consumer = new FlinkKinesisConsumer<>(
    streamName,
    // block 2
    GlueSchemaRegistryAvroDeserializationSchema.forGeneric(schema, configs),
    properties);

FlinkKinesisProducer<GenericRecord> producer = new FlinkKinesisProducer<>(
    // block 3
    GlueSchemaRegistryAvroSerializationSchema.forGeneric(schema, topic, configs),
    properties);
producer.setDefaultStream(streamName);
producer.setDefaultPartition("0");

DataStream<GenericRecord> stream = env.addSource(consumer);
stream.addSink(producer);
env.execute();

Use Case: Integration with Amazon Lambda

To use an Amazon Lambdafunction as an Apache Kafka/Amazon MSK consumer and deserialize Avro-encoded messages using Amazon Glue Schema Registry, visit the MSK Labs page.

Use case: Amazon Glue Data Catalog

Amazon Glue tables support schemas that you can specify manually or by reference to the Amazon Glue Schema Registry. The Schema Registry integrates with the Data Catalog to allow you to optionally use schemas stored in the Schema Registry when creating or updating Amazon Glue tables or partitions in the Data Catalog. To identify a schema definition in the Schema Registry, at a minimum, you need to know the ARN of the schema it is part of. A schema version of a schema, which contains a schema definition, can be referenced by its UUID or version number. There is always one schema version, the "latest" version, that can be looked up without knowing its version number or UUID.

When calling the CreateTable or UpdateTable operations, you will pass a TableInput structure that contains a StorageDescriptor, which may have a SchemaReference to an existing schema in the Schema Registry. Similarly, when you call the GetTable or GetPartition APIs, the response may contain the schema and the SchemaReference. When a table or partition was created using a schema references, the Data Catalog will try to fetch the schema for this schema reference. In case it is unable to find the schema in the Schema Registry, it returns an empty schema in the GetTable response; otherwise the response will have both the schema and schema reference.

You can also perform the actions from the Amazon Glue console.

To perform these operations and create, update, or view the schema information, you must give an IAM role to the calling user that provides permissions for the GetSchemaVersion API.

Adding a table or updating the schema for a table

Adding a new table from an existing schema binds the table to a specific schema version. Once new schema versions get registered, you can update this table definition from the View table page in the Amazon Glue console or using the UpdateTable action (Python: update_table) API.

Adding a table from an existing schema

You can create an Amazon Glue table from a schema version in the registry using the Amazon Glue console or CreateTable API.

Amazon Glue API

When calling the CreateTable API, you will pass a TableInput that contains a StorageDescriptor which has a SchemaReference to an existing schema in the Schema Registry.

Amazon Glue console

To create a table from the Amazon Glue console:

The schema and version applied to the table appears in the Glue schema column in the list of tables. You can view the table to see more details.

Updating the schema for a table

When a new schema version becomes available, you may want to update a table's schema using the UpdateTable action (Python: update_table) API or the Amazon Glue console.

Amazon Glue API

When calling the UpdateTable API, you will pass a TableInput that contains a StorageDescriptor which has a SchemaReference to an existing schema in the Schema Registry.

Amazon Glue console

To update the schema for a table from the Amazon Glue console:

Use case: Amazon Glue streaming

Amazon Glue streaming consumes data from streaming sources and perform ETL operations before writing to an output sink. Input streaming source can be specified using a Data Table or directly by specifying the source configuration.

Amazon Glue streaming supports a Data Catalog table for the streaming source created with the schema present in the Amazon Glue Schema Registry. You can create a schema in the Amazon Glue Schema Registry and create an Amazon Glue table with a streaming source using this schema. This Amazon Glue table can be used as an input to an Amazon Glue streaming job for deserializing data in the input stream.

One point to note here is when the schema in the Amazon Glue Schema Registry changes, you need to restart the Amazon Glue streaming job needs to reflect the changes in the schema.

Use case: Apache Kafka Streams

The Apache Kafka Streams API is a client library for processing and analyzing data stored in Apache Kafka. This section describes the integration of Apache Kafka Streams with Amazon Glue Schema Registry, which allows you to manage and enforce schemas on your data streaming applications. For more information on Apache Kafka Streams, see Apache Kafka Streams.

Integrating with the SerDes Libraries

There is a GlueSchemaRegistryKafkaStreamsSerde class that you can configure a Streams application with.

Kafka Streams application example code

To use the Amazon Glue Schema Registry within an Apache Kafka Streams application:

Implementation results

These results show the filtering process of records that were filtered out in step 3 as a favorite_color of "pink" or value of "15.0".

Records before filtering:

{"name": "Sansa", "favorite_number": 99, "favorite_color": "white"}
{"name": "Harry", "favorite_number": 10, "favorite_color": "black"}
{"name": "Hermione", "favorite_number": 1, "favorite_color": "red"}
{"name": "Ron", "favorite_number": 0, "favorite_color": "pink"}
{"name": "Jay", "favorite_number": 0, "favorite_color": "pink"}
	
{"id": "commute_1","amount": 3.5}
{"id": "grocery_1","amount": 25.5}
{"id": "entertainment_1","amount": 19.2}
{"id": "entertainment_2","amount": 105}
	{"id": "commute_1","amount": 15}
	

Records after filtering:

{"name": "Sansa", "favorite_number": 99, "favorite_color": "white"}
{"name": "Harry", "favorite_number": 10, "favorite_color": "black"}
{"name": "Hermione", "favorite_number": 1, "favorite_color": "red"}
{"name": "Ron", "favorite_number": 0, "favorite_color": "pink"}
	
{"id": "commute_1","amount": 3.5}
{"id": "grocery_1","amount": 25.5}
{"id": "entertainment_1","amount": 19.2}
{"id": "entertainment_2","amount": 105}
	

Use case: Apache Kafka Connect

The integration of Apache Kafka Connect with the Amazon Glue Schema Registry enables you to get schema information from connectors. The Apache Kafka converters specify the format of data within Apache Kafka and how to translate it into Apache Kafka Connect data. Every Apache Kafka Connect user will need to configure these converters based on the format they want their data in when loaded from or stored into Apache Kafka. In this way, you can define your own converters to translate Apache Kafka Connect data into the type used in the Amazon Glue Schema Registry (for example: Avro) and utilize our serializer to register its schema and do serialization. Then converters are also able to use our deserializer to deserialize data received from Apache Kafka and convert it back into Apache Kafka Connect data. An example workflow diagram is given below.