KafkaEx

KafkaEx is an Elixir client for Apache Kafka with support for Kafka versions 0.8.0 and newer. KafkaEx requires Elixir 1.6+ and Erlang OTP 19+.

See http://hexdocs.pm/kafka_ex/ for documentation, https://github.com/kafkaex/kafka_ex/ for code.

KafkaEx supports the following Kafka features:

• Broker and Topic Metadata
• Produce Messages
• Fetch Messages
• Message Compression with Snappy and gzip
• Offset Management (fetch / commit / autocommit)
• Consumer Groups
• Topics Management (create / delete)

See Kafka Protocol Documentation and A Guide to the Kafka Protocol for details of these features.

IMPORTANT - Kayrock and The Future of KafkaEx

TL;DR:

• This is new implementation and we need people to test it!
• Set kafka_version: "kayrock" to use the new client implementation.
• The new client should be compatible with existing code when used this way.
• Many functions now support an api_version parameter, see below for details, e.g., how to store offsets in Kafka instead of Zookeeper.
• Version 1.0 of KafkaEx will be based on Kayrock and have a cleaner API - you can start testing this API by using modules from the KafkaEx.New namespace. See below for details.
• Version 0.11.0+ of KafkaEx is required to use Kayrock.

To support some oft-requested features (offset storage in Kafka, message timestamps), we have integrated KafkaEx with Kayrock which is a library that handles serialization and deserialization of the Kafka message protocol in a way that can grow as Kafka does.

Unfortunately, the existing KafkaEx API is built in such a way that it doesn't easily support this growth. This, combined with a number of other existing warts in the current API, has led us to the conclusion that v1.0 of KafkaEx should have a new and cleaner API.

The path we have planned to get to v1.0 is:

1. Add a Kayrock compatibility layer for the existing KafkaEx API (DONE, not released).
2. Expose Kayrock's API versioning through a select handful of KafkaEx API functions so that users can get access to the most-requested features (e.g., offset storage in Kafka and message timestamps) (DONE, not released).
3. Begin designing and implementing the new API in parallel in the KafkaEx.New namespace (EARLY PROGRESS).
4. Incrementally release the new API alongside the legacy API so that early adopters can test it.
5. Once the new API is complete and stable, move it to the KafkaEx namespace (i.e., drop the New part) and it will replace the legacy API. This will be released as v1.0.

Users of KafkaEx can help a lot by testing the new code. At first, we need people to test the Kayrock-based client using compatibility mode. You can do this by simply setting kafka_version: "kayrock" in your configuration. That should be all you need to change. If you want to test new features enabled by api_versions options then that is also very valuable to us (see below for links to details). Then, as work on the new API ramps up, users can contribute feedback to pull requests (or even contribute pull requests!) and test out the new API as it becomes available.

For more information on using the Kayrock-based client, see

• Github: kayrock.md
• HexDocs: kayrock-based client

For more information on the v1.0 API, see

• Github: new_api.md
• HexDocs: New API

Using KafkaEx in an Elixir project

The standard approach for adding dependencies to an Elixir application applies: add KafkaEx to the deps list in your project's mix.exs file. You may also optionally add snappyer (required only if you want to use snappy compression).

# mix.exs
defmodule MyApp.Mixfile do
  # ...

  defp deps do
    [
      # add to your existing deps
      {:kafka_ex, "~> 0.11"},
      # If using snappy-erlang-nif (snappy) compression
      {:snappy, git: "https://github.com/fdmanana/snappy-erlang-nif"}
      # if using snappyer (snappy) compression
      {:snappyer, "~> 1.2"}
    ]
  end
end

Then run mix deps.get to fetch dependencies.

Configuration

See config/config.exs or KafkaEx.Config for a description of configuration variables, including the Kafka broker list and default consumer group.

You can also override options when creating a worker, see below.

Timeouts with SSL

When using certain versions of OTP, random timeouts can occur if using SSL.

Impacted versions:

• OTP 21.3.8.1 -> 21.3.8.14
• OTP 22.1 -> 22.3.1

Upgrade respectively to 21.3.8.15 or 22.3.2 to solve this.

Usage Examples

Consumer Groups

To use a consumer group, first implement a handler module using KafkaEx.GenConsumer.

defmodule ExampleGenConsumer do
  use KafkaEx.GenConsumer

  alias KafkaEx.Protocol.Fetch.Message

  require Logger

  # note - messages are delivered in batches
  def handle_message_set(message_set, state) do
    for %Message{value: message} <- message_set do
      Logger.debug(fn -> "message: " <> inspect(message) end)
    end
    {:async_commit, state}
  end
end

Then add a KafkaEx.ConsumerGroup to your application's supervision tree and configure it to use the implementation module.

See the KafkaEx.GenConsumer and KafkaEx.ConsumerGroup documentation for details.

Create a KafkaEx Worker

KafkaEx worker processes manage the state of the connection to the Kafka broker.

iex> KafkaEx.create_worker(:pr) # where :pr is the process name of the created worker
{:ok, #PID<0.171.0>}

With custom options:

iex> uris = [{"localhost", 9092}, {"localhost", 9093}, {"localhost", 9094}]
[{"localhost", 9092}, {"localhost", 9093}, {"localhost", 9094}]
iex> KafkaEx.create_worker(:pr, [uris: uris, consumer_group: "kafka_ex", consumer_group_update_interval: 100])
{:ok, #PID<0.172.0>}

Create an unnamed KafkaEx worker

You may find you want to create many workers, say in conjunction with a poolboy pool. In this scenario you usually won't want to name these worker processes.

To create an unnamed worked with create_worker:

iex> KafkaEx.create_worker(:no_name) # indicates to the server process not to name the process
{:ok, #PID<0.171.0>}

Use KafkaEx with a pooling library

Note that KafkaEx has a supervisor to manage its workers. If you are using Poolboy or a similar library, you will want to manually create a worker so that it is not supervised by KafkaEx.Supervisor. To do this, you will need to call:

GenServer.start_link(KafkaEx.Config.server_impl,
  [
    [uris: KafkaEx.Config.brokers(),
     consumer_group: Application.get_env(:kafka_ex, :consumer_group)],
    :no_name
  ]
)

Alternatively, you can call

KafkaEx.start_link_worker(:no_name)

Retrieve kafka metadata

For all metadata

iex> KafkaEx.metadata
%KafkaEx.Protocol.Metadata.Response{brokers: [%KafkaEx.Protocol.Metadata.Broker{host:
 "192.168.59.103",
   node_id: 49162, port: 49162, socket: nil}],
 topic_metadatas: [%KafkaEx.Protocol.Metadata.TopicMetadata{error_code: :no_error,
   partition_metadatas: [%KafkaEx.Protocol.Metadata.PartitionMetadata{error_code: :no_error,
     isrs: [49162], leader: 49162, partition_id: 0, replicas: [49162]}],
   topic: "LRCYFQDVWUFEIUCCTFGP"},
  %KafkaEx.Protocol.Metadata.TopicMetadata{error_code: :no_error,
   partition_metadatas: [%KafkaEx.Protocol.Metadata.PartitionMetadata{error_code: :no_error,
     isrs: [49162], leader: 49162, partition_id: 0, replicas: [49162]}],
   topic: "JSIMKCLQYTWXMSIGESYL"},
  %KafkaEx.Protocol.Metadata.TopicMetadata{error_code: :no_error,
   partition_metadatas: [%KafkaEx.Protocol.Metadata.PartitionMetadata{error_code: :no_error,
     isrs: [49162], leader: 49162, partition_id: 0, replicas: [49162]}],
   topic: "SCFRRXXLDFPOWSPQQMSD"},
  %KafkaEx.Protocol.Metadata.TopicMetadata{error_code: :no_error,
...

For a specific topic

iex> KafkaEx.metadata(topic: "foo")
%KafkaEx.Protocol.Metadata.Response{brokers: [%KafkaEx.Protocol.Metadata.Broker{host: "192.168.59.103",
   node_id: 49162, port: 49162, socket: nil}],
 topic_metadatas: [%KafkaEx.Protocol.Metadata.TopicMetadata{error_code: :no_error,
   partition_metadatas: [%KafkaEx.Protocol.Metadata.PartitionMetadata{error_code: :no_error,
     isrs: [49162], leader: 49162, partition_id: 0, replicas: [49162]}],
   topic: "foo"}]}

Retrieve offset from a particular time

Kafka will get the starting offset of the log segment that is created no later than the given timestamp. Due to this, and since the offset request is served only at segment granularity, the offset fetch request returns less accurate results for larger segment sizes.

iex> KafkaEx.offset("foo", 0, {{2015, 3, 29}, {23, 56, 40}}) # Note that the time specified should match/be ahead of time on the server that kafka runs
[%KafkaEx.Protocol.Offset.Response{partition_offsets: [%{error_code: :no_error, offset: [256], partition: 0}], topic: "foo"}]

Retrieve the latest offset

iex> KafkaEx.latest_offset("foo", 0) # where 0 is the partition
[%KafkaEx.Protocol.Offset.Response{partition_offsets: [%{error_code: :no_error, offset: [16], partition: 0}], topic: "foo"}]

Retrieve the earliest offset

iex> KafkaEx.earliest_offset("foo", 0) # where 0 is the partition
[%KafkaEx.Protocol.Offset.Response{partition_offsets: [%{error_code: :no_error, offset: [0], partition: 0}], topic: "foo"}]

Fetch kafka logs

NOTE You must pass auto_commit: false in the options for fetch/3 when using Kafka < 0.8.2 or when using :no_consumer_group.

iex> KafkaEx.fetch("foo", 0, offset: 5) # where 0 is the partition and 5 is the offset we want to start fetching from
[%KafkaEx.Protocol.Fetch.Response{partitions: [%{error_code: :no_error,
     hw_mark_offset: 115,
     message_set: [
      %KafkaEx.Protocol.Fetch.Message{attributes: 0, crc: 4264455069, key: nil, offset: 5, value: "hey"},
      %KafkaEx.Protocol.Fetch.Message{attributes: 0, crc: 4264455069, key: nil, offset: 6, value: "hey"},
      %KafkaEx.Protocol.Fetch.Message{attributes: 0, crc: 4264455069, key: nil, offset: 7, value: "hey"},
      %KafkaEx.Protocol.Fetch.Message{attributes: 0, crc: 4264455069, key: nil, offset: 8, value: "hey"},
      %KafkaEx.Protocol.Fetch.Message{attributes: 0, crc: 4264455069, key: nil, offset: 9, value: "hey"}
...], partition: 0}], topic: "foo"}]

Produce kafka logs

iex> KafkaEx.produce("foo", 0, "hey") # where "foo" is the topic and "hey" is the message
:ok

Stream kafka logs

See the KafkaEx.stream/3 documentation for details on streaming.

iex> KafkaEx.produce("foo", 0, "hey")
:ok
iex> KafkaEx.produce("foo", 0, "hi")
:ok
iex> KafkaEx.stream("foo", 0, offset: 0) |> Enum.take(2)
[%{attributes: 0, crc: 4264455069, key: nil, offset: 0, value: "hey"},
 %{attributes: 0, crc: 4251893211, key: nil, offset: 1, value: "hi"}]

For Kafka < 0.8.2 the stream/3 requires auto_commit: false

iex> KafkaEx.stream("foo", 0, offset: 0, auto_commit: false) |> Enum.take(2)

Compression

Snappy and gzip compression is supported. Example usage for producing compressed messages:

message1 = %KafkaEx.Protocol.Produce.Message{value: "value 1"}
message2 = %KafkaEx.Protocol.Produce.Message{key: "key 2", value: "value 2"}
messages = [message1, message2]

#snappy
produce_request = %KafkaEx.Protocol.Produce.Request{
  topic: "test_topic",
  partition: 0,
  required_acks: 1,
  compression: :snappy,
  messages: messages}
KafkaEx.produce(produce_request)

#gzip
produce_request = %KafkaEx.Protocol.Produce.Request{
  topic: "test_topic",
  partition: 0,
  required_acks: 1,
  compression: :gzip,
  messages: messages}
KafkaEx.produce(produce_request)

Compression is handled automatically on the consuming/fetching end.

Testing

It is strongly recommended to test using the Dockerized test cluster described below. This is required for contributions to KafkaEx.

NOTE You may have to run the test suite twice to get tests to pass. Due to asynchronous issues, the test suite sometimes fails on the first try.

Dockerized Test Cluster

Testing KafkaEx requires a local SSL-enabled Kafka cluster with 3 nodes: one node listening on each port 9092, 9093, and 9093. The easiest way to do this is using the scripts in this repository that utilize Docker and Docker Compose (both of which are freely available). This is the method we use for our CI testing of KafkaEx.

To launch the included test cluster, run

./scripts/docker_up.sh

The docker_up.sh script will attempt to determine an IP address for your computer on an active network interface.

The test cluster runs Kafka 0.11.0.1.

Running the KafkaEx Tests

The KafkaEx tests are split up using tags to handle testing multiple scenarios and Kafka versions.

Unit tests

These tests do not require a Kafka cluster to be running (see test/test_helper.exs:3 for the tags excluded when running this).

mix test --no-start

Integration tests

If you are not using the Docker test cluster, you may need to modify config/config.exs for your set up.

The full test suite requires Kafka 2.1.0+.

Kafka >= 0.9.0

The 0.9 client includes functionality that cannot be tested with older clusters.

./scripts/all_tests.sh

Kafka = 0.9.0

The 0.9 client includes functionality that cannot be tested with older clusters.

mix test --include integration --include consumer_group --include server_0_p_9_p_0

Kafka >= 0.8.2 and < 0.9.0

Kafka 0.8.2 introduced the consumer group API.

mix test --include consumer_group --include integration

Kafka < 0.8.2

If your test cluster is older, the consumer group tests must be omitted.

mix test --include integration --include server_0_p_8_p_0

Static analysis

mix dialyzer

Contributing

All contributions are managed through the kafkaex github repo.

If you find a bug or would like to contribute, please open an issue or submit a pull request. Please refer to CONTRIBUTING.md for our contribution process.

KafkaEx has a Slack channel: #kafkaex on elixir-lang.slack.com. You can request an invite via http://bit.ly/slackelixir. The Slack channel is appropriate for quick questions or general design discussions. The Slack discussion is archived at http://slack.elixirhq.com/kafkaex.

Default snappy algorithm use snappyer package

It can be changed to snappy by using this:

config :kafka_ex, snappy_module: :snappy

Snappy erlang nif is deprecated and will be dropped 1.0.0 release.