前因

在之前的篇幅中--EFK 配置geo-ip落地实践，已经介绍过如何配置geo插件到fluentd中，实现ip地址转经纬度的实例。

但是随之也产生了一个问题。即，ip出口地址大多数落在了大城市，并且相对集中，其导致根据其转换的经纬度所绘制的地图点太集中，不能够反映真实的用户所在位置。

因此，新的需求从天而降！我们需要使用客户端上报的经纬度来绘制地图。

在浏览了所有fluentd的插件后。居然没有解析request body的插件。 fluentd插件查询地址：https://www.fluentd.org/plugins/all

没办法，自己动手写一个。

插件编写 首先，我们可以明确，fluentd的filter数据是顺序流向下一个的。由于我也是第一次编写Ruby程序。所以先看一下项目结构：

其中比较重要的是：

filter_parse_request_body.rb （fluentd的filter插件编写，用于拦截数据，中途修改数据）
out_parse_request_body.rb （fluentd的out插件编写，用于最后格式化输出数据）
parse_request_body_extractor.rb （最终执行拆解数据的类）
fluent-plugin-parse_request_body.gemspec （项目的配置文件，包括项目名称，版本号等）

下面，直接上代码：

filter_parse_request_body.rb

require "fluent/plugin/filter"
require "fluent/plugin/parse_request_body_extractor"

module Fluent::Plugin
  class ParseRequestBodyFilter < Fluent::Plugin::Filter

    #注册filter名称#
    Fluent::Plugin.register_filter('parse_request_body', self)

    #request body在record中所对应的key#
    desc "point a key whose value contains body string."
    config_param :key,    :string
    #需要使用array_value替换数据的key#
    desc "point a key who will be replaced."
    config_param :replace_key,    :string, default: nil
    #需要被组合在一起的数据名称#
    desc "If set, the key/value will be reformd array whose would be added to the record."
    config_param :array_value,   :string, default: nil
    #最终组合数据在record中的key#
    desc "array_value's key in record"
    config_param :array_value_key,   :string, default: nil
    #request body解析数据白名单#
    desc "If set, only the key/value whose key is included only will be added to the record."
    config_param :only,   :string, default: nil
    #request body解析数据黑名单#
    desc "If set, the key/value whose key is included except will NOT be added to the record."
    config_param :except, :string, default: nil
    #是否只保留最终解析出来的数据，而删除request body原数据#
    desc "If set to true, the original key url will be discarded from the record."
    config_param :discard_key, :bool, default: false
    #给解析出的数据key添加前缀#
    desc "Prefix of fields."
    config_param :add_field_prefix, :string, default: nil
    #是否允许解析空key#
    desc "If set to true, permit blank key."
    config_param :permit_blank_key, :bool, default: false

    #初始化解析器#
    def configure(conf)
      super
      @extractor = Fluent::Plugin::ParseRequestBodyExtractor.new(self, conf)
    end

    #执行解析工作#
    def filter(tag, time, record)
      @extractor.add_query_params_field(record)
    end
  end
end

out_parse_request_body.rb （fluentd的out插件编写，用于最后格式化输出数据）

require "fluent/plugin/output"
require "fluent/plugin/parse_request_body_extractor"

module Fluent::Plugin
  class ParseRequestBodyOutput < Fluent::Plugin::Output
    include Fluent::HandleTagNameMixin

    #注册output名称#
    Fluent::Plugin.register_output('parse_request_body', self)

    helpers :event_emitter

    #request body在record中所对应的key#
    desc "point a key whose value contains body string."
    config_param :key,    :string
    #需要使用array_value替换数据的key#
    desc "point a key who will be replaced."
    config_param :replace_key,    :string, default: nil
    #需要被组合在一起的数据名称#
    desc "If set, the key/value will be reformd array whose would be added to the record."
    config_param :array_value,   :string, default: nil
    #最终组合数据在record中的key#
    desc "array_value's key in record"
    config_param :array_value_key,   :string, default: nil
    #request body解析数据白名单#
    desc "If set, only the key/value whose key is included only will be added to the record."
    config_param :only,   :string, default: nil
    #request body解析数据黑名单#
    desc "If set, the key/value whose key is included except will NOT be added to the record."
    config_param :except, :string, default: nil
    #是否只保留最终解析出来的数据，而删除request body原数据#
    desc "If set to true, the original key url will be discarded from the record."
    config_param :discard_key, :bool, default: false
    #给解析出的数据key添加前缀#
    desc "Prefix of fields."
    config_param :add_field_prefix, :string, default: nil
    #是否允许解析空key#
    desc "If set to true, permit blank key."
    config_param :permit_blank_key, :bool, default: false

    #初始化解析器#
    def configure(conf)
      super
      @extractor = Fluent::Plugin::ParseRequestBodyExtractor.new(self, conf)
    end

    def multi_workers_ready?
      true
    end

    #执行解析工作#
    def filter_record(tag, time, record)
      record = @extractor.add_query_params_field(record)
      super(tag, time, record)
    end

    def process(tag, es)
      es.each do |time, record|
        t = tag.dup
        filter_record(t, time, record)
        router.emit(t, time, record)
      end
    end
  end
end

parse_request_body_extractor.rb （最终执行拆解数据的类）

require 'uri'
require 'cgi/util'
require 'webrick'

module Fluent::Plugin
  class ParseRequestBodyExtractor

    attr_reader :log
        
    #初始化解析器#
    def initialize(plugin, conf)
      @log = plugin.log

      if plugin.is_a?(Fluent::Plugin::Output)
        unless have_tag_option?(plugin)
          raise Fluent::ConfigError, "out_parse_request_body: At least one of remove_tag_prefix/remove_tag_suffix/add_tag_prefix/add_tag_suffix is required to be set."
        end
      end

      #从配置中读取配置选项#
      @key = plugin.key
      @only = plugin.only
      @except = plugin.except
      @discard_key = plugin.discard_key
      @add_field_prefix = plugin.add_field_prefix
      @permit_blank_key = plugin.permit_blank_key
      @array_value = plugin.array_value
      @array_value_key = plugin.array_value_key
      @replace_key = plugin.replace_key

      #初始化白名单#
      if @only
        @include_keys = @only.split(/\s*,\s*/).inject({}) do |hash, i|
          hash[i] = true
          hash
        end
      end

      #初始化黑名单#
      if @except
        @exclude_keys = @except.split(/\s*,\s*/).inject({}) do |hash, i|
          hash[i] = true
          hash
        end
      end

      #初始化需要被组合的key#
      if @array_value_key
        if @array_value
          @include_array_value = @array_value.split(/\s*,\s*/).inject({}) do |hash, i|
            hash[i] = true
            hash
          end
        end
      end

    end

    #解析方法#
    def add_query_params_field(record)
      return record unless record[@key]
      add_query_params(record[@key], record)
      replace_record_by_key(record) if @replace_key
      record.delete(@key) if @discard_key
      record
    end

    private

    #替换record中某一个键值#
    def replace_record_by_key(record)
      return record unless record[@replace_key]
      replace_value = record[@array_value_key]
      empty_value = replace_value.select {|item| item == 0 }
      if replace_value && (empty_value.size != replace_value.size)
        record[@replace_key] = replace_value
      end
    end

    def have_tag_option?(plugin)
      plugin.remove_tag_prefix ||
        plugin.remove_tag_suffix ||
        plugin.add_tag_prefix    ||
        plugin.add_tag_suffix
    end

    def create_field_key(field_key)
      if add_field_prefix?
        "#{@add_field_prefix}#{field_key}"
      else
        field_key
      end
    end

    def add_field_prefix?
      !!@add_field_prefix
    end

    def permit_blank_key?
      @permit_blank_key
    end

    def add_query_params(body, record)
      return if body.nil?
      placeholder = []
      body.split('&').each do |pair|
        key, value = pair.split('=', 2).map { |i| CGI.unescape(i) }
        next if (key.nil? || key.empty?) && (!permit_blank_key? || value.nil? || value.empty?)
        key ||= ''
        value ||= ''

        new_key = create_field_key(key)
        if @only
          record[new_key] = value if @include_keys.has_key?(key)
        elsif @except
          record[new_key] = value if !@exclude_keys.has_key?(key)
        else
          record[new_key] = value
        end

        if @include_array_value
          placeholder[placeholder.size] = value.to_f if @include_array_value.has_key?(key)
        end
      end

      unless placeholder.empty?
        record[@array_value_key] = placeholder
      end
    end
  end
end

fluent-plugin-parse_request_body.gemspec （项目的配置文件，包括项目名称，版本号等）

Gem::Specification.new do |gem|
  gem.name          = 'fluent-plugin-parse_request_body'
  gem.version       = '0.0.18'
  gem.authors       = ['EkiSong']
  gem.email         = ['yifriday0614@gmail.com']
  gem.homepage      = 'https://github.com/yifriday/fluent-plugin-parse_request_body.git'
  gem.description   = %q{Fluentd plugin to parse request body.}
  gem.summary       = %q{Fluentd plugin to parse request body}
  gem.license       = 'MIT'

  gem.files         = `git ls-files`.split($\)
  gem.executables   = gem.files.grep(%r{^bin/}).map{ |f| File.basename(f) }
  gem.test_files    = gem.files.grep(%r{^(test|spec|features)/})
  gem.require_paths = ['lib']

  if defined?(RUBY_VERSION) && RUBY_VERSION > '2.2'
    gem.add_development_dependency "test-unit", '~> 3'
  end

  gem.add_development_dependency 'rake'
  gem.add_development_dependency 'appraisal'
  gem.add_runtime_dependency     'fluentd', ['>= 0.14.8', '< 2']
end

##插件制作

在结束编码工作之后，使用gem制作上传插件。

ruby所写的插件，都可以上传到这个网站https://rubygems.org/

下面是在命令行中的相关操作

➜  fluent-plugin-parse_request_body git:(master) gem build fluent-plugin-parse_request_body.gemspec 
  Successfully built RubyGem
  Name: fluent-plugin-parse_request_body
  Version: 0.0.18
  File: fluent-plugin-parse_request_body-0.0.18.gem
➜  fluent-plugin-parse_request_body git:(master) gem push fluent-plugin-parse_request_body-0.0.18.gem
Pushing gem to https://rubygems.org...
Successfully registered gem: fluent-plugin-parse_request_body (0.0.18)

在成功上传插件后。我们就可以使用gem工具，在我们的服务器中安装了

插件安装和配置

安装直接执行下面语句即可：

/usr/sbin/td-agent-gem install fluent-plugin-parse_request_body

而在td-agent.conf中的配置如下：

<filter nginx.**>
  @type geoip
    geoip_lookup_key remote
    geoip_database /opt/td-agent/embedded/lib/ruby/gems/2.1.0/gems/fluent-plugin-geoip-1.2.0/data/GeoLiteCity.dat
    geoip2_database /opt/td-agent/embedded/lib/ruby/gems/2.1.0/gems/fluent-plugin-geoip-1.2.0/data/GeoLite2-City.mmdb
    <record>
     city         ${city.names.zh-CN["remote"]}     # skip adding fields if this field is null
     country      ${country.iso_code["remote"]}
     country_name ${country.names.zh-CN["remote"]}
     location '[${location.longitude["remote"]},${location.latitude["remote"]}]'
    </record>
</filter>

<filter nginx.**>
  @type parse_request_body
  key                request_body
  #discard_key        true
  add_field_prefix   body.
  only lat,lng
  array_value lat,lng
  array_value_key locatonGPS
  replace_key location
</filter>

注意：由于我们需要在没有客户端经纬度的情况下，依然使用ip转换，并且filter是单向数据流，所以 parse_request_body一定要配置在geo下方。

至此，数据地图的数据处理部分第二阶段工作结束。

后续，在目前的需求上，此插件需要优化一点：

组合数据，使用标签动态配置，想geo中配置location一样，而不是现在死配置这种比较low的做法。

感兴趣的同学，可以看我的github：

https://github.com/yifriday/fluent-plugin-parse_request_body.git