Kafka Connect for Azure Cosmos DB (SQL API)

Introduction

File any issues / feature requests / questions etc. you may have in the Issues for this repo.

This project provides connectors for Kafka Connect to read from and write data to Azure Cosmos DB(SQL API).

The connectors in this repository are specifically for the Cosmos DB SQL API. If you are using Cosmos DB with other APIs then there is likely a specific connector for that API, but it's not this one.

• Mongo API - MongoDB Kafka Connector
• Cassandra API - Cassandra Sink Connector
• Gremlin API - Kafka Connect connector for Cosmos DB Gremlin API

Exactly-Once Support

• Source Connector
  • For the time being, this connector supports at-least once with multiple tasks and exactly-once for single tasks.
• Sink Connector
  • The sink connector fully supports exactly-once semantics.

Supported Data Formats

The sink & source connectors are configurable in order to support:

Since key and value settings, including the format and serialization, can be independently configured in Kafka, it is possible to work with different data formats for records' keys and values respectively.

To cater for this there is converter configuration for both key.converter and value.converter.

Converter Configuration Examples

JSON (Plain)

• If you need to use JSON without Schema Registry for Connect data, you can use the JsonConverter supported with Kafka. The example below shows the JsonConverter key and value properties that are added to the configuration:

  key.converter=org.apache.kafka.connect.json.JsonConverter
  key.converter.schemas.enable=false
  value.converter=org.apache.kafka.connect.json.JsonConverter
  value.converter.schemas.enable=false
  

JSON with Schema

• When the properties key.converter.schemas.enable and value.converter.schemas.enable are set to true, the key or value is not treated as plain JSON, but rather as a composite JSON object containing both an internal schema and the data.

  key.converter=org.apache.kafka.connect.json.JsonConverter
  key.converter.schemas.enable=true
  value.converter=org.apache.kafka.connect.json.JsonConverter
  value.converter.schemas.enable=true
  

• The resulting message to Kafka would look like the example below, with schema and payload top-level elements in the JSON:

  {
    "schema": {
      "type": "struct",
      "fields": [
        {
          "type": "int32",
          "optional": false,
          "field": "userid"
        },
        {
          "type": "string",
          "optional": false,
          "field": "name"
        }
      ],
      "optional": false,
      "name": "ksql.users"
    },
    "payload": {
      "userid": 123,
      "name": "user's name"
    }
  }
  

NOTE: The message written is made up of the schema + payload. Notice the size of the message, as well as the proportion of it that is made up of the payload vs. the schema. This is repeated in every message you write to Kafka. In scenarios like this, you may want to use a serialisation format like JSON Schema or Avro, where the schema is stored separately and the message holds just the payload.

AVRO

• This connector supports AVRO. To use AVRO you need to configure a AvroConverter so that Kafka Connect knows how to work with AVRO data. This connector has been tested with the AvroConverter supplied by Confluent, under Apache 2.0 license, but another custom converter can be used in its place instead if you prefer.

• Kafka deals with keys and values independently, you need to specify the key.converter and value.converter properties as required in the worker configuration.

• An additional converter property must also be added, when using AvroConverter, that provides the URL for the Schema Registry.

The example below shows the AvroConverter key and value properties that are added to the configuration:

key.converter=io.confluent.connect.avro.AvroConverter
key.converter.schema.registry.url=http://schema-registry:8081
value.converter=io.confluent.connect.avro.AvroConverter
value.converter.schema.registry.url=http://schema-registry:8081

Choosing a conversion format

• If you're configuring a Source connector and

  • If you want Kafka Connect to include plain JSON in the message it writes to Kafka, you'd set JSON (Plain) configuration.
  • If you want Kafka Connect to include the schema in the message it writes to Kafka, you’d set JSON with Schema configuration.
  • If you want Kafka Connect to include AVRO format in the message it writes to Kafka, you'd set AVRO configuration.

• If you’re consuming JSON data from a Kafka topic in to a Sink connector, you need to understand how the JSON was serialised when it was written to the Kafka topic:

  • If it was with JSON serialiser, then you need to set Kafka Connect to use the JSON converter (org.apache.kafka.connect.json.JsonConverter).
    • If the JSON data was written as a plain string, then you need to determine if the data includes a nested schema/payload. If it does,then you would set, JSON with Schema configuration.
    • However, if you’re consuming JSON data and it doesn’t have the schema/payload construct, then you must tell Kafka Connect not to look for a schema by setting schemas.enable=false as per JSON (Plain) configuration.
  • If it was with AVRO serialiser, then you need to set Kafka Connect to use the AVRO converter (io.confluent.connect.avro.AvroConverter) as per AVRO configuration.

Common Errors

Some of the common errors you can get if you misconfigure the converters in Kafka Connect. These will show up in the sinks you configure for Kafka Connect, as it’s this point at which you’ll be trying to deserialize the messages already stored in Kafka. Converter problems tend not to occur in sources because it’s in the source that the serialization is set.

Configuration

Common Configuration Properties

The Sink and Source connectors share the following common configuration properties

For Sink connector specific configuration, please refer to the Sink Connector Documentation

For Source connector specific configuration, please refer to the Source Connector Documentation

Project Setup

Please refer Developer Walkthrough and Project Setup for initial setup instructions.

Performance Testing

For more information on the performance tests run for the Sink and Source connectors, refer to the Performance testing document.

Refer to the Performance Environment Setup for exact steps on deploying the performance test environment for the Connectors.

Dead Letter Queue

We introduced the standard dead level queue from Kafka. For more info see: https://www.confluent.io/blog/kafka-connect-deep-dive-error-handling-dead-letter-queues

Resources

• Kafka Connect
• Kafka Connect Deep Dive – Converters and Serialization Explained