物联网架构成长之路(8)-EMQ-Hook了解、连接Kafka发送消息

1. 前言

　　按照我自己设计的物联网框架，对于MQTT集群中的所有消息，是要持久化到磁盘的，这里采用一个消息队列中间件Kafka作为数据缓冲，缓冲结果存到数据仓库中，以供后续作为数据分析。由于MQTT集群中的消息都是比较分散的，所以使用Kafka来聚合、采集消息。

2. 下载&编译&安装

　　Kafka依赖ZooKeeper

　　在这里下载 http://mirrors.shuosc.org/apache/zookeeper/zookeeper-3.4.11/zookeeper-3.4.11.tar.gz

　　http://mirrors.shuosc.org/apache/kafka/1.0.0/kafka-1.0.0-src.tgz

　　http://mirror.bit.edu.cn/apache/kafka/1.0.0/kafka_2.12-1.0.0.tgz

　　学习的话， 可以参考这个文档 http://orchome.com/kafka/index

　　配置使用过程可以参考官网的 　http://kafka.apache.org/quickstart  http://zookeeper.apache.org/doc/current/zookeeperStarted.html ，有些资料因为版本升级的原因，已经不是以前的那种启动方式了。https://cwiki.apache.org/confluence/display/KAFKA/Clients

3. 启动Zookeeper

 cp ./conf/zoo_sample.cfg ./conf/zoo.cfg
 ./bin/zkServer.sh start

　　上图，就表示启动单机模式。

　　./bin/zkCli.sh -server 127.0.0.1:2181 进行连接

　　然后进入命令行模式，可以慢慢玩了

　　单击模式就没有什么要配置的。最多修改zoo.cfg中的dataDir文件

　　ZooKeeper启动replicated模式，集群模式

　　zoo.cfg 增加服务器集群信息

 server.=172.16.20.229::
 server.=172.16.23.203::
 server.=172.16.23.204::

 ./bin/zkServer.sh start-foreground #启动

　　注意在echo “1” > %dataDir%/myid 对于每个服务器都要创建一个myid文件

　　启动都是会有一些奇奇怪怪的问题，上网找资料就可以了。

　　一般第一台ZooKeeper启动是会有Connection refused 出错，这个是正常的，后面的两台还没有启动，不过后面也一个一个启动了。

　　如果过程中，有一个断开了，然后修改数据，然后这个断开的又连上了，那么ZooKeeper集群内部会镜像diff

 -- ::, [myid:] - INFO  [QuorumPeer[myid=]/0.0.0.0::Learner@] - Getting a diff from the leader 0x100000005

　　然后就用客户端A

 ./bin/zkCli.sh -server 172.16.20.229:
 ls /
 create /zk_test my_data
 ls /

　　然后用客户端B

 ./bin/zkCli.sh -server 172.16.23.203:
 ls /
 get /zk_test
 ls /

　　可以看到ZooKeeper信息在内部进行了共享

　　具体可以参考这篇博客 http://www.cnblogs.com/sunddenly/p/4018459.html

4. 启动 Kafka

　　由于kafka以来ZooKeeper，所以有了上面一步的ZooKeeper了解。

　　实际中，可以直接下载kafka的二进制包，直接使用，http://mirror.bit.edu.cn/apache/kafka/1.0.0/kafka_2.12-1.0.0.tgz

　　启动Zookeeper

 ./bin/zookeeper-server-start.sh config/zookeeper.properties

　　启动Kafka

 ./bin/kafka-server-start.sh config/server.properties

　　创建主题

 ./bin/kafka-topics.sh --create --zookeeper localhost: --replication-factor  --partitions  --topic test

　　查看主题

 ./bin/kafka-topics.sh --list --zookeeper localhost:

　　发送消息

 ./bin/kafka-console-producer.sh --broker-list localhost: --topic test

　　消费消息

 ./bin/kafka-console-consumer.sh --zookeeper localhost: --topic test --from-beginning

5. Erlang 连接 kafka

　　主要参考这个 https://github.com/msdevanms/emqttd_plugin_kafka_bridge

　　增加依赖 https://github.com/helpshift/ekaf.git

　　首先在Makefile增加

 DEPS = eredis ecpool clique ekaf
 dep_ekaf = git https://github.com/helpshift/ekaf master

　　然后在rebar.config 增加

 {ekaf, “.*”, {git, “https://github.com/helpshift/ekaf”, “master”}}

　　在etc/emq_plugin_wunaozai.conf 中增加

 ## kafka config
 wunaozai.msg.kafka.server = 127.0.0.1:
 wunaozai.msg.kafka.topic = test

　　在priv/emq_plugin_wunaozai.schema 中增加

 %% wunaozai.msg.kafka.server = 127.0.0.1:
 {
     mapping,
     "wunaozai.msg.kafka.server",
     "emq_plugin_wunaozai.kafka",
     [
         {default, {"127.0.0.1", }},
         {datatype, [integer, ip, string]}
     ]
 }.

 %% wunaozai.msg.kafka.topic = test
 {
     mapping,
     "wunaozai.msg.kafka.topic",
     "emq_plugin_wunaozai.kafka",
     [
         {default, "test"},
         {datatype, string},
         hidden
     ]
 }.

 %% translation
 {
     translation,
     "emq_plugin_wunaozai.kafka",
     fun(Conf) ->
             {RHost, RPort} = case cuttlefish:conf_get("wunaozai.msg.kafka.server", Conf) of
                                  {Ip, Port} -> {Ip, Port};
                                  S          -> case string:tokens(S, ":") of
                                                    [Domain]       -> {Domain, };
                                                    [Domain, Port] -> {Domain, list_to_integer(Port)}
                                                end
                              end,
             Topic = cuttlefish:conf_get("wunaozai.msg.kafka.topic", Conf),
             [
              {host, RHost},
              {port, RPort},
              {topic, Topic}
             ]
     end
 }.

6. 数据发往Kafka

　　接下来，由于功能基本上是基于EMQ框架的Hook钩子设计，在EMQ接收到客户端上下线、主题订阅或消息发布确认时，触发钩子顺序执行回调函数，所以大部分功能在 src/emq_plugin_wunaozai.erl 文件进行修改。

 -module(emq_plugin_wunaozai).

 -include("emq_plugin_wunaozai.hrl").

 -include_lib("emqttd/include/emqttd.hrl").

 -export([load/, unload/]).

 %% Hooks functions

 -export([on_client_connected/, on_client_disconnected/]).

 -export([on_client_subscribe/, on_client_unsubscribe/]).

 -export([on_session_created/, on_session_subscribed/, on_session_unsubscribed/, on_session_terminated/]).

 -export([on_message_publish/, on_message_delivered/, on_message_acked/]).

 %% Called when the plugin application start
 load(Env) ->
     ekaf_init([Env]),
     emqttd:hook('client.connected', fun ?MODULE:on_client_connected/, [Env]),
     emqttd:hook('client.disconnected', fun ?MODULE:on_client_disconnected/, [Env]),
     emqttd:hook('client.subscribe', fun ?MODULE:on_client_subscribe/, [Env]),
     emqttd:hook('client.unsubscribe', fun ?MODULE:on_client_unsubscribe/, [Env]),
     emqttd:hook('session.created', fun ?MODULE:on_session_created/, [Env]),
     emqttd:hook('session.subscribed', fun ?MODULE:on_session_subscribed/, [Env]),
     emqttd:hook('session.unsubscribed', fun ?MODULE:on_session_unsubscribed/, [Env]),
     emqttd:hook('session.terminated', fun ?MODULE:on_session_terminated/, [Env]),
     emqttd:hook('message.publish', fun ?MODULE:on_message_publish/, [Env]),
     emqttd:hook('message.delivered', fun ?MODULE:on_message_delivered/, [Env]),
     emqttd:hook('message.acked', fun ?MODULE:on_message_acked/, [Env]),
     io:format("start wunaozai Test Reload.~n", []).

 on_client_connected(ConnAck, Client = #mqtt_client{client_id = ClientId}, _Env) ->
     io:format("client ~s connected, connack: ~w~n", [ClientId, ConnAck]),
     ekaf_send(<<"connected">>, ClientId, {}, _Env),
     {ok, Client}.

 on_client_disconnected(Reason, _Client = #mqtt_client{client_id = ClientId}, _Env) ->
     io:format("client ~s disconnected, reason: ~w~n", [ClientId, Reason]),
     ekaf_send(<<"disconnected">>, ClientId, {}, _Env),
     ok.

 on_client_subscribe(ClientId, Username, TopicTable, _Env) ->
     io:format("client(~s/~s) will subscribe: ~p~n", [Username, ClientId, TopicTable]),
     {ok, TopicTable}.

 on_client_unsubscribe(ClientId, Username, TopicTable, _Env) ->
     io:format("client(~s/~s) unsubscribe ~p~n", [ClientId, Username, TopicTable]),
     {ok, TopicTable}.

 on_session_created(ClientId, Username, _Env) ->
     io:format("session(~s/~s) created.", [ClientId, Username]).

 on_session_subscribed(ClientId, Username, {Topic, Opts}, _Env) ->
     io:format("session(~s/~s) subscribed: ~p~n", [Username, ClientId, {Topic, Opts}]),
     ekaf_send(<<"subscribed">>, ClientId, {Topic, Opts}, _Env),
     {ok, {Topic, Opts}}.

 on_session_unsubscribed(ClientId, Username, {Topic, Opts}, _Env) ->
     io:format("session(~s/~s) unsubscribed: ~p~n", [Username, ClientId, {Topic, Opts}]),
     ekaf_send(<<"unsubscribed">>, ClientId, {Topic, Opts}, _Env),
     ok.

 on_session_terminated(ClientId, Username, Reason, _Env) ->
     io:format("session(~s/~s) terminated: ~p.~n", [ClientId, Username, Reason]),
     stop.

 %% transform message and return
 on_message_publish(Message = #mqtt_message{topic = <<"$SYS/", _/binary>>}, _Env) ->
     {ok, Message};
 on_message_publish(Message, _Env) ->
     io:format("publish ~s~n", [emqttd_message:format(Message)]),
     ekaf_send(<<"public">>, {}, Message, _Env),
     {ok, Message}.

 on_message_delivered(ClientId, Username, Message, _Env) ->
     io:format("delivered to client(~s/~s): ~s~n", [Username, ClientId, emqttd_message:format(Message)]),
     {ok, Message}.

 on_message_acked(ClientId, Username, Message, _Env) ->
     io:format("client(~s/~s) acked: ~s~n", [Username, ClientId, emqttd_message:format(Message)]),
     {ok, Message}.

 %% Called when the plugin application stop
 unload() ->
     emqttd:unhook('client.connected', fun ?MODULE:on_client_connected/),
     emqttd:unhook('client.disconnected', fun ?MODULE:on_client_disconnected/),
     emqttd:unhook('client.subscribe', fun ?MODULE:on_client_subscribe/),
     emqttd:unhook('client.unsubscribe', fun ?MODULE:on_client_unsubscribe/),
     emqttd:unhook('session.created', fun ?MODULE:on_session_created/),
     emqttd:unhook('session.subscribed', fun ?MODULE:on_session_subscribed/),
     emqttd:unhook('session.unsubscribed', fun ?MODULE:on_session_unsubscribed/),
     emqttd:unhook('session.terminated', fun ?MODULE:on_session_terminated/),
     emqttd:unhook('message.publish', fun ?MODULE:on_message_publish/),
     emqttd:unhook('message.delivered', fun ?MODULE:on_message_delivered/),
     emqttd:unhook('message.acked', fun ?MODULE:on_message_acked/).

 %% ==================== ekaf_init STA.===============================%%
 ekaf_init(_Env) ->
     % clique 方式读取配置文件
     Env = application:get_env(?APP, kafka),
     {ok, Kafka} = Env,
     Host = proplists:get_value(host, Kafka),
     Port = proplists:get_value(port, Kafka),
     Broker = {Host, Port},
     Topic = proplists:get_value(topic, Kafka),
     io:format("~w ~w ~w ~n", [Host, Port, Topic]),

     % init kafka
     application:set_env(ekaf, ekaf_partition_strategy, strict_round_robin),
     application:set_env(ekaf, ekaf_bootstrap_broker, Broker),
     application:set_env(ekaf, ekaf_bootstrap_topics, list_to_binary(Topic)),
     %application:set_env(ekaf, ekaf_bootstrap_broker, {"127.0.0.1", }),
     %application:set_env(ekaf, ekaf_bootstrap_topics, <<"test">>),

     %io:format("Init ekaf with ~s:~b~n", [Host, Port]),
     %%ekaf:produce_async_batched(<<"test">>, list_to_binary(Json)),
     ok.
 %% ==================== ekaf_init END.===============================%%

 %% ==================== ekaf_send STA.===============================%%
 ekaf_send(Type, ClientId, {}, _Env) ->
     Json = mochijson2:encode([
                               {type, Type},
                               {client_id, ClientId},
                               {message, {}},
                               {cluster_node, node()},
                               {ts, emqttd_time:now_ms()}
                              ]),
     ekaf_send_sync(Json);
 ekaf_send(Type, ClientId, {Reason}, _Env) ->
     Json = mochijson2:encode([
                               {type, Type},
                               {client_id, ClientId},
                               {cluster_node, node()},
                               {message, Reason},
                               {ts, emqttd_time:now_ms()}
                              ]),
     ekaf_send_sync(Json);
 ekaf_send(Type, ClientId, {Topic, Opts}, _Env) ->
     Json = mochijson2:encode([
                               {type, Type},
                               {client_id, ClientId},
                               {cluster_node, node()},
                               {message, [
                                          {topic, Topic},
                                          {opts, Opts}
                                         ]},
                               {ts, emqttd_time:now_ms()}
                              ]),
     ekaf_send_sync(Json);
 ekaf_send(Type, _, Message, _Env) ->
     Id = Message#mqtt_message.id,
     From = Message#mqtt_message.from, %需要登录和不需要登录这里的返回值是不一样的
     Topic = Message#mqtt_message.topic,
     Payload = Message#mqtt_message.payload,
     Qos = Message#mqtt_message.qos,
     Dup = Message#mqtt_message.dup,
     Retain = Message#mqtt_message.retain,
     Timestamp = Message#mqtt_message.timestamp,

     ClientId = c(From),
     Username = u(From),

     Json = mochijson2:encode([
                               {type, Type},
                               {client_id, ClientId},
                               {message, [
                                          {username, Username},
                                          {topic, Topic},
                                          {payload, Payload},
                                          {qos, i(Qos)},
                                          {dup, i(Dup)},
                                          {retain, i(Retain)}
                                         ]},
                               {cluster_node, node()},
                               {ts, emqttd_time:now_ms()}
                              ]),
     ekaf_send_sync(Json).

 ekaf_send_async(Msg) ->
     Topic = ekaf_get_topic(),
     ekaf_send_async(Topic, Msg).
 ekaf_send_async(Topic, Msg) ->
     ekaf:produce_async_batched(list_to_binary(Topic), list_to_binary(Msg)).
 ekaf_send_sync(Msg) ->
     Topic = ekaf_get_topic(),
     ekaf_send_sync(Topic, Msg).
 ekaf_send_sync(Topic, Msg) ->
     ekaf:produce_sync_batched(list_to_binary(Topic), list_to_binary(Msg)).

 i(true) -> ;
 i(false) -> ;
 i(I) when is_integer(I) -> I.
 c({ClientId, Username}) -> ClientId;
 c(From) -> From.
 u({ClientId, Username}) -> Username;
 u(From) -> From.
 %% ==================== ekaf_send END.===============================%%

 %% ==================== ekaf_set_host STA.===============================%%
 ekaf_set_host(Host) ->
     ekaf_set_host(Host, ).
 ekaf_set_host(Host, Port) ->
     Broker = {Host, Port},
     application:set_env(ekaf, ekaf_bootstrap_broker, Broker),
     io:format("reset ekaf Broker ~s:~b ~n", [Host, Port]),
     ok.
 %% ==================== ekaf_set_host END.===============================%%

 %% ==================== ekaf_set_topic STA.===============================%%
 ekaf_set_topic(Topic) ->
     application:set_env(ekaf, ekaf_bootstrap_topics, list_to_binary(Topic)),
     ok.
 ekaf_get_topic() ->
     Env = application:get_env(?APP, kafka),
     {ok, Kafka} = Env,
     Topic = proplists:get_value(topic, Kafka),
     Topic.
 %% ==================== ekaf_set_topic END.===============================%%

　　上面是所有源代码，下面对其进行简单说明

　　ekaf_init 函数，主要对配置文件的读取和解析并存放到application的环境变量中

　　ekaf_send 函数，主要是封装成对应的JSON数据，然后发到Kafka中

　　ekaf_send_async 函数，主要是异步发送JSON数据，不确保发往Kafka的顺序与Kafka消费者的接收时的顺序

　　ekaf_send_sync 函数，是同步发送JSON数据，确保按照顺序发往kafka与Kafka消费者有序接收数据

　　ekaf_set_host 函数，设置kafka的域名与端口

　　ekaf_set_topic 函数，设置发往kafka时的主题

　　ekaf_get_topic 函数，获取当前主题

　　load函数增加ekaf_init调用

　　剩下的在每个钩子回调中调用 ekaf_send函数

7. 测试

　　（1）启动Zookeeper ./bin/zookeeper-server-start.sh config/zookeeper.properties

　　（2）启动Kafka ./bin/kafka-server-start.sh config/server.properties

　　（3）启动消费者 ./bin/kafka-console-consumer.sh --zookeeper localhost:2181 --topic test --from-beginning

　　（4）启动一个生产者 ./bin/kafka-console-producer.sh --broker-list localhost:9092 --topic test

　　（5）启动EMQ ./_rel/emqttd/bin/emqttd console

　　（6）打开MQTT客户端并连接、订阅、发布等操作

　　（7）可以在消费者界面上看到获取到的信息

8. 插件源码

　　最后给出本次插件开发的所有源码

　　https://files.cnblogs.com/files/wunaozai/emq_plugin_wunaozai.zip