How to Write Filter Plugin

This section shows how to write a custom filter plugin in addition to the core ones. The plugin filenames, starting with filter_ prefix, are registered as filter plugins.

See Plugin Base Class API for more details on the common APIs of all the plugins.

Here is the implementation of the most basic filter that passes through all the events as-is:

require 'fluent/plugin/filter'

module Fluent::Plugin
  class PassThruFilter < Filter
    # Register this filter as "passthru"
    Fluent::Plugin.register_filter('passthru', self)

    # config_param works like other plugins

    def configure(conf)
      super
      # Do the usual configuration here
    end

    # def start
    #   super
    #   # Override this method if anything needed as startup.
    # end

    # def shutdown
    #   # Override this method to use it to free up resources, etc.
    #   super
    # end

    def filter(tag, time, record)
      # Since our example is a pass-thru filter, it does nothing and just
      # returns the record as-is.
      # If returns nil, that records are ignored.
      record
    end
  end
end

Methods

A filter plugin overrides the one of filter/filter_with_time/filter_stream method.

#filter(tag, time, record)

This method implements the filtering logic.

  • tag: is a String,

  • time is a Fluent::EventTime or an Integer; and,

  • record is a Hash with String keys.

The return value of this method should be a Hash of modified record, or nil. If it is nil, the event will be discarded.

# example
def filter(tag, time, record)
  # process record
  record['fluentd_tag'] = tag
  record
end

#filter_with_time(tag, time, record)

This method implements the filtering logic with time update. Event time will be replaced with the return value.

  • tag: is a String,

  • time is a Fluent::EventTime or an Integer; and,

  • record is a Hash with String keys.

The return value of this method should be two element array, [new_time, new_record] , or nil. If it is nil, the event will be discarded.

# example
def filter_with_time(tag, time, record)
  new_time = get_time_from_record(record)
  new_record = update_record(tag, record)
  return new_time, new_record  # this is same with return [new_time, new_record]
end

#filter_stream(tag, es)

This method implements the event stream based filtering logic. If you hard to implement the logic with filter, e.g. need to handle multiple records in one processing, use this method.

The return value of this method should be MultiEventStream. If it is nil, the event will be discarded.

# example
def filter_stream(tag, es)
  new_es = Fluent::MultiEventStream.new
  es.each { |time, record|
    new_time = process_time(tag, time, record)
    new_record = process_record(tag, time, record)
    new_es.add(time, record)
  }
  new_es
end

Writing Tests

Fluentd filter plugin has one or some points to be tested. Others (parsing configurations, controlling buffers, retries, flushes and many others) are controlled by Fluentd core.

Fluentd also provides test driver for plugins. You can write tests for your own plugins very easily:

# test/plugin/test_filter_your_own.rb

require 'test/unit'
require 'fluent/plugin/test/driver/filter'

# your own plugin
require 'fluent/plugin/filter_your_own'

class YourOwnFilterTest < Test::Unit::TestCase
  def setup
    Fluent::Test.setup # this is required to setup router and others
  end

  # default configuration for tests
  CONFIG = %[
    param1 value1
    param2 value2
  ]

  def create_driver(conf = CONFIG)
    Fluent::Test::Driver::Filter.new(Fluent::Plugin::YourOwnFilter).configure(conf)
  end

  def filter(config, messages)
    d = create_driver(config)
    d.run(default_tag: 'input.access') do
      messages.each do |message|
        d.feed(message)
      end
    end
    d.filtered_records
  end

  sub_test_case 'configured with invalid configuration' do
    test 'empty configuration' do
      assert_raise(Fluent::ConfigError) do
         create_driver('')
      end
    end

    test 'param1 should reject too short string' do
      conf = %[
        param1 a
      ]
      assert_raise(Fluent::ConfigError) do
         create_driver(conf)
      end
    end
    # ...
  end

  sub_test_case 'plugin will add some fields' do
    test 'add hostname to record' do
      conf = CONFIG
      messages = [
        { 'message' => 'This is test message' }
      ]
      expected = [
        { 'message' => 'This is test message', 'hostname' => 'example.com' }
      ]
      filtered_records = filter(conf, messages)
      assert_equal(expected, filtered_records)
    end
    # ...
  end
  # ...
end

Overview of Tests

Testing for the filter plugins is mainly for:

  • Validation of configuration parameters (i.e. #configure)

  • Validation of the filtered records

To make testing easy, the plugin test driver provides a dummy router, a logger and general functionality to override the system, parser and other relevant configurations.

The lifecycle of the plugin and its test driver is:

  1. Instantiate the test driver which then instantiates the plugin

  2. Configure plugin

  3. Register conditions to stop/break running tests

  4. Run test code (provided as a block for d.run)

  5. Assert results of tests using data provided by the driver

At the start of Step # 4, the test driver calls the startup methods of the plugin e.g. #start and at the end #stop, #shutdown, etc. It can be skipped by optional arguments of #run.

For:

  • configuration tests, repeat steps # 1-2

  • full feature tests, repeat steps # 1-5

For more details, see Testing API for Plugins.

If this article is incorrect or outdated, or omits critical information, please let us know. Fluentd is an open-source project under Cloud Native Computing Foundation (CNCF). All components are available under the Apache 2 License.

Last updated