A Kafka client for Erlang and Elixir

Copyright (c) 2014, 2015 Finexkap, 2015, 2016, 2017, 2018 G-Corp, 2015, 2016, 2017 BotsUnit

Version: 2.2.0

Authors: Gregoire Lejeune (gregoire.lejeune@finexkap.com), Gregoire Lejeune (greg@g-corp.io), Gregoire Lejeune (gregoire.lejeune@botsunit.com).

Version 2.0.0 cause changes in the following APIs :

• kafe:start_consumer/3

• kafe:fetch/3

Kafe has been tested with Kafka 0.9 and above

You can also use it with Kafka 0.8 but kafe_consumer is not compatible with this version.

Links

• Apache Kafka
• Apache Kafka Protocol

Configuration

brokers	[{inet:hostname(), inet:port_number()}]	List of brokers	[{"localhost", 9092}]
pool_size	integer()	Initial connection pool/brocker	5
chunk_pool_size	integer()	Size of new connection pool/brocker	10
brokers_update_frequency	integer()	Frequency (ms) for brokers update	60000
protocol_timeout	integer()	API call timeout	60000
client_id	binary()	Client ID Name	<<"kafe">>
api_version	[{integer(), integer()}] | integer() | auto	API Version	auto*
correlation_id	integer()	Correlation ID	0
socket	[{sndbuf, integer()}, {recbuf, integer()}, {buffer, integer()}]	Socker configuration	[{sndbuf, 4194304}, {recbuf, 4194304}, {buffer, 4194304}]

^* use 0 with Kafka >= 0.8 < 0.9 ; auto with Kafka >= 0.9 < 0.10

Example :

[
  {kafe, [
    {brokers, [
      {"localhost", 9092},
      {"localhost", 9093},
      {"localhost", 9094}
    ]},
    {pool_size, 1},
    {chunk_pool_size, 2},
    {brokers_update_frequency, 10000},
    {client_id, <<"kafe">>},
    {api_version, auto},
    {correlation_id, 0},
    {socket, [
      {sndbuf, 4194304},
      {recbuf, 4194304},
      {buffer, 4194304}
    ]},
  ]}
]

Kafe use lager ; see also how to configure it.

Custom API version

If you need to, you can specify each version of the protocol APIs. To do so, use a list as value for the api_version's configuration. This list must be a list of tuples where the first element is the api key and the second is the api version.

Example :

...
{api_version, [{0, 0}, {0, 1}, {0, 0}, ...]}
...

Create a consumer

Using a function

To create a consumer, create a function with 6 parameters :

-module(my_consumer).

-export([consume/6]).

consume(CommitID, Topic, Partition, Offset, Key, Value) ->
  % Do something with Topic/Partition/Offset/Key/Value
  ok.

The consume function must return ok if the message was treated, or {error, term()} on error.

Then start a new consumer :

...
kafe:start(),
...
kafe:start_consumer(my_group, fun my_consumer:consume/6, Options),
...

See kafe:start_consumer/3 for the available Options.

In the consume function, if you didn't start the consumer in autocommit mode (using before_processing | after_processing in the commit options), you need to commit manually when you have finished to treat the message. To do so, use kafe_consumer:commit/4.

When you are done with your consumer, stop it :

...
kafe:stop_consumer(my_group),
...

Using the kafe_consumer_subscriber behaviour

-module(my_consumer).
-behaviour(kafe_consumer_subscriber).

-export([init/4, handle_message/2]).
-include_lib("kafe/include/kafe_consumer.hrl").

-record(state, {
               }).

init(Group, Topic, Partition, Args) ->
  % Do something with Group, Topic, Partition, Args
  {ok, #state{}}.

handle_message(Message, State) ->
  % Do something with Message
  % And update your State (if needed)
  {ok, NewState}.

Then start a new consumer :

...
kafe:start().
...
kafe:start_consumer(my_group, {my_consumer, Args}, Options).
% Or
kafe:start_consumer(my_group, my_consumer, Options).
...

To commit a message (if you need to), use kafe_consumer:commit/4.

Using with Elixir

Elixir' users can use Kafe and Kafe.Consumer instead of :kafe and :kafe_consumer.

defmodule My.Consumer do
  def consume(commit_id, topic, partition, offset, key, value) do
    # Do something with topic/partition/offset/key/value
    :ok
  end
end

defmodule My.Consumer.Subscriber do
  @behaviour Kafe.Consumer.Subscriber
  require Kafe.Records

  def init(group, topic, partition, args) do
    % Do something with group/topic/partition/args
    % and create the state
    {:ok, state}
  end

  def handle_message(message, state) do
    msg = Kafe.Records.message(message)
    % Do something with msg and update (or not) the state
    {:ok, new_state}
  end
end

...
Kafe.start()
...
Kafe.start_consumer(:my_group, &My.Consumer.consume/6, options)
# or
Kafe.start_consumer(:my_group, {My.Consumer.Subscriber, args}, options)
# or
Kafe.start_consumer(:my_group, My.Consumer.Subscriber, options)
...
Kafe.stop_consumer(:my_group)
...

Metrics

To enable metrics :

1/ Add metrics > 2.2 in your dependencies.

2/ Set enable_metrics to true in the kafe configuration :

{kafe, [
  ...
  {enable_metrics, true},
  ...
]}

3/ Adding a metrics module in your configuration :

{metrics, [
  {metrics_mod, metrics_folsom}
]}

You can choose between Folsom ({metrics_mod, metrics_folsom}), Exometer ({metrics_mod, metrics_exometer}) or Grapherl ({metrics_mod, metrics_grapherl}).

Be sure that's Folsom, Exometer or Grapherl and metrics is started before starting Kafe.

application:ensure_all_started(folsom).
application:ensure_all_started(metrics).
application:ensure_all_started(kafe).

Metrics are disabled by default.

Kafe offers the following metrics :

Name	Type	Description
kafe_consumer.CONSUMER_GROUP.messages.fetch	gauge	Number of received messages on the last fetch for the CONSUMER_GROUP
kafe_consumer.CONSUMER_GROUP.TOPIC.PARTITION.messages.fetch	gauge	Number of received messages on the last fetch for the {TOPIC, PARTITION} and CONSUMER_GROUP
kafe_consumer.CONSUMER_GROUP.messages	counter	Total number of received messages for the CONSUMER_GROUP
kafe_consumer.CONSUMER_GROUP.TOPIC.PARTITION.messages	counter	Total number of received messages for the {TOPIC, PARTITION} and CONSUMER_GROUP
kafe_consumer.CONSUMER_GROUP.TOPIC.PARTITION.duration.fetch	gauge	Fetch duration (ms) per message, for the {TOPIC, PARTITION} and CONSUMER_GROUP
kafe_consumer.CONSUMER_GROUP.TOPIC.PARTITION.pending_commits	gauge	Number of pending commits, for the {TOPIC, PARTITION} and CONSUMER_GROUP

You can add a prefix to all metrics by adding a metrics_prefix in the metrics configuration :

{metrics, [
  {metrics_mod, metrics_folsom},
  {metrics_prefix, my_bot}
]}

Build and tests

Kafe use rebar3 and bu.mk. So, you can use :

• ./rebar3 compile to compile Kafe.

• ./rebar3 eunit to run tests.

• ./rebar3 ct to run (integration) tests.

• ./rebar3 edoc to build documentation.

• ./rebar3 elixir generate_mix to generate mix.exs file.

• ./rebar3 elixir generate_lib to generate Elixir bindings.

Or

• make release Tag and release to hex.pm

• make integ Run integration tests

• make docker-compose.yml Create docker-compose.yml

• make docker-start Start docker

• make docker-stop Stop docker

• make elixir Generate Elixir bindings (mix.exs and libs)

• make tests Run tests

• make doc Generate doc

• make dist Create a distribution

• make clean Clean

• make distclean Clean the distribution

• make info Display application informations

• make tag Create a git tag

• make local.hex Install hexfor Mix

• make local.rebar Install rebar for Mix

• make bu-mk Update bu.mk

• make help Show this help.

To run the integration tests, you must start zookeeper and a kafka cluster (3 brokers) and have the following three topics :

• testone : replication factor: 1, partitions: 1

• testtwo : replication factor: 2, partitions: 2

• testthree : replication factor: 3, partitions: 3

You can use the makefile rules docker-compose.yml and docker-start to help you to create this environment using docker (tested on Linux only).

API Documentation

See documentation

Contributing

1. Fork it ( https://github.com/G-Corp/kafe/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request

Licence

kafe is available for use under the following license, commonly known as the 3-clause (or "modified") BSD license:

Copyright (c) 2014, 2015 Finexkap
Copyright (c) 2015, 2016, 2017 BotsUnit
Copyright (c) 2015, 2016, 2017, 2018 G-Corp

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
• The name of the author may not be used to endorse or promote products derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE AUTHOR AS IS AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Modules

kafe

kafe_consumer

kafe_consumer_subscriber