ELK的架构原理：

logstash收集nginx日志，并对日志进行过滤拆分，并将处理后的结构化数据输出给elastcsearch，es对日志进行存储和索引构建，kibana提供图形界面及对es 查询api进行了封装，提供友好的查询和统计页面。
在生产环境中，logstash作为agent安装部署在任何想要收集日志的主机上，为了缓解多个agent对ES的输出压力，需要定义一个broker（redis）对日志进行输入缓冲，然后定义一个logstash server对broker中的日志统一读取并输出给ES集群。broker常常使用redis，为了broker的高可用，还可以对redis做集群部署。
单点安装测试只部署一个es，一个logstash agent，一个kibana，一个nginx。

安装测试流程：

1.安装nginx-1.12.0

#安装gcc等编译工具

sudo yum groupinstall -y '开发工具'

#安装nginx需要的pcre ，zlib开发库

yum install -y pcre-devel zlib-devel  

#创建nginx的安装目录

mkdir nginx

#配置编译安装nginx
tar zxf nginx-1.12.0.tar.gz

cd nginx-1.12.0

./configure --prefix=/home/hoewon/nginx 

make 

make install

#简单配置nginx

user root

#运行
sudo nginx

2.安装logstash

#
tar zxf logstash-5.5.2.tar.gz

#对grok-pattern做连接

ln -s $LOGSTASH_HOME/ vendor/bundle/jruby/1.9/gems/logstash-patterns-core-4.1.1/patterns/grok-patterns/grok-patterns grok-patterns

#在grok-patterns追加nginx日志的模式，因为对http_x_forwarded_for 的匹配不好使，所以zhushidiaole

NGUSER %{NGUSERNAME}
NGINXACCESS %{IPORHOST:clientip} - %{NOTSPACE:remote_user} \[%{HTTPDATE:timestamp}\] \"(?:%{WORD:verb} %{NOTSPACE:request}(?: HTTP/%{NUMBER:httpversion})?|%{DATA:rawrequest})\" %{NUMBER:response} (?:%{NUMBER:bytes}|-) %{QS:referrer} %{QS:agent}
# %{NOTSPACE:http_x_forwarded_for}

#编辑logstash启动脚本
vim simple.conf

input {
        file{
                path => ["/home/hoewon/nginx/logs/access.log"]
                type => "nginxlog"
                start_position => "beginning"
        }

}

filter{
        grok{
                match => {

                        "message" => "%{NGINXACCESS}"
                }
        }
}

output{
        stdout{
                codec => rubydebug
        }

}

#检查运行配置文件
 bin/logstash -t -f simple.conf

#运行logstash，并测试输出
 bin/logstash  -f simple.conf

输出如下

{
        "request" => "/favicon.ico",
          "agent" => "\"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.104 Safari/537.36 Core/1.53.3368.400 QQBrowser/9.6.11974.400\"",
           "verb" => "GET",
        "message" => "192.168.247.1 - - [08/Sep/2017:15:25:46 +0800] \"GET /favicon.ico HTTP/1.1\" 403 571 \"-\" \"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.104 Safari/537.36 Core/1.53.3368.400 QQBrowser/9.6.11974.400\"",
           "type" => "nginxlog",
    "remote_user" => "-",
           "path" => "/home/hoewon/nginx/logs/access.log",
       "referrer" => "\"-\"",
     "@timestamp" => 2017-09-08T08:04:19.534Z,
       "response" => "403",
          "bytes" => "571",
       "clientip" => "192.168.247.1",
       "@version" => "1",
           "host" => "kube01",
    "httpversion" => "1.1",
      "timestamp" => "08/Sep/2017:15:25:46 +0800"
}

测试输入输出无误的话，更改output插件为elsaticsearch

input {
        file{
                path => ["/home/hoewon/nginx/logs/access.log"]
                type => "nginxlog"
                start_position => "beginning"
        }

}

filter{
        grok{
                match => {

                        "message" => "%{NGINXACCESS}"
                }
        }
}

output{

        elasticsearch{
                hosts => ["192.168.247.142:9200"]
                index => "nginxlog"

        }
}

3.安装Elasticsearch

# 
tar zxf elasticsearch-5.5.2.tar.gz

#
sudo vim /etc/security/limits.conf

#<domain>        <type>  <item>  <value>
hoewon           soft    nofile     65536
hoewon           hard    nofile     65536
hoewon           soft    nproc      2048
hoewon           hard    nproc      2048

#modify the vm.max_map_count

sudo vim /etc/sysctl.conf

  vm.max_map_count=262144

#
sysctl -p

#vim $ES_HOME/conf/elasticsearch.conf
network.host:    192.168.247.142 （or 0.0.0.0）
http.port: port
#如果是集群修改如下配置，集群是通过cluster.name自动在9300端口上寻找节点信息的
node.name: nodename
cluster.name: clustername

#
$ES_HOME/bin/elasticsearch

4.安装kibana

#
tar zxf kibana-5.5.2-linux-x86_64.tar.gz

# 
vim $KIBANA_HOME/conf/kibana.yml

server.host: "192.168.247.142"
elasticsearch.url: "http://192.168.247.142:9200"

#
$KIBANA_HOME/bin/kibana

测试：
访问nginx所在主机80端口。logstash会自动收集日志，并输出给es，登录kibana所在主机:5601/，配置好es index的pattern，然后在discover中就可以查到文档信息。如下：