nginx系列之五: 負載均衡

**

前言

**

nginx系列之一:nginx入門
nginx系列之二:配置文件解讀
nginx系列之三:日誌配置
nginx系列之四:web服務器
nginx系列之五: 負載均衡
nginx系列之六:cache服務
nginx系列之七:限流配置
nginx系列之八:使用upsync模塊實現負載均衡

轉自:在此感謝原博主的整理分享

使用nginx做負載均衡的兩大模塊:

  • upstream 定義負載節點池。
  • location 模塊 進行URL匹配。
  • proxy模塊 發送請求給upstream定義的節點池。

一、upstream模塊解讀

nginx 的負載均衡功能依賴於 ngx_http_upstream_module模塊,所支持的代理方式有 proxy_pass(一般用於反向代理),fastcgi_pass(一般用於和動態程序交互),memcached_pass,proxy_next_upstream,fastcgi_next_pass,memcached_next_pass 。

upstream 模塊應該放於http{}標籤內。

模塊寫法:

upstream backend {
    ip_hash; 
    server backend1.example.com       weight=5;
    server backend2.example.com:8080;
    server backup1.example.com:8080   backup;
    server backup2.example.com:8080   backup;
}
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7

實例一:

upstream dynamic {
    zone upstream_dynamic 64k;
server backend1.example.com      weight=5;
server backend2.example.com:8080 fail_timeout=5s slow_start=30s;
server 192.0.2.1                 max_fails=3;
server backend3.example.com      resolve;

server backup1.example.com:8080  backup;
server backup2.example.com:8080  backup;

}

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11

語法解釋:

1.1 nginx默認支持四種調度算法

  • 輪詢(rr),每個請求按時間順序逐一分配到不同的後端服務器,如果後端服務器故障,故障系統自動清除,使用戶訪問不受影響。
  • 輪詢權值(weight),weight值越大,分配到的訪問機率越高,主要用於後端每個服務器性能不均的情況。
  • ip_hash,每個請求按訪問IP的hash結果分配,這樣來自同一個IP的固定訪問一個後端服務器,主要解決動態網站session共享的問題。
  • url_hash,按照訪問的URL的hash結果來分配請求,是每個URL定向到同一個後端服務器,可以進一步提高後端緩存服務器的效率,nginx本身不支持,如果想使用需要安裝nginx的hash軟件包。
  • fair,這個算法可以依據頁面大小和加載時間長短智能的進行負載均衡,也就是根據後端服務器的響應時間來分配請求,相應時間短的優先分配,默認不支持,如果想使用需要安裝upstream_fail模塊。
  • least_conn 最少鏈接數,那個機器連接數少就分發。

1.2 server模塊的寫法

server IP 調度狀態
  • 1

server指令指定後端服務器IP地址和端口,同時還可以設定每個後端服務器在負載均衡調度中的狀態。

  • down 表示當前的server暫時不參與負載均衡。
  • backup 預留的備份服務器,當其他所有的非backup服務器出現故障或者忙的時候,纔會請求backup機器,因爲這臺集羣的壓力最小。
  • max_fails 允許請求失敗的次數,默認是1,當超過最大次數時,返回proxy_next_upstream模塊定義的錯誤。0表示禁止失敗嘗試,企業場景:2-3.京東1次,藍汛10次,根據業務需求去配置。
  • fail_timeout,在經歷了max_fails次失敗後,暫停服務的時間。京東是3s,藍汛是3s,根據業務需求配置。常規業務2-3秒合理。

例:如果max_fails是5,他就檢測5次,如果五次都是502.那麼,他就會根據fail_timeout 的值,等待10秒,再去檢測。

server 如果接域名,需要內網有DNS服務器,或者在負載均衡器的hosts文件做域名解析。server後面還可以直接接IP或IP加端口。

1.3 長連接 keepalive

upstream backend {
    server backend2.example.com:8080;
    server backup1.example.com:8080   backup;
    keepalive 100;
}
  • 1
  • 2
  • 3
  • 4
  • 5

通過該指令配置了每個worker進程與上游服務器可緩存的空閒連接的最大數量。
當超出這個數量時,最近最少使用的連接將被關閉。keepalive指令不限制worker進程與上游服務器的總連接。

location / {
    # 支持keep-alive
    proxy_http_version 1.1;
    proxy_set_header Connection "";
    proxy_pass http://backup;
}
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 如果是http/1.0 需要配置發送"Connection: Keep-Alive" 請求頭。
  • 上游服務器不要忘記開啓長連接支持。

連接池配置建議

  • 總長連接數是"空閒連接池"+"釋放連接池"的長連接總數。
  • 首先,長連接配置不會限制worker進程可以打開的總連接數(超了的作爲短連接)。另外連接池一定要根據場景合理進行設置。
  1. 空閒連接池太小,連接不夠用,需要不斷建連接。
  2. 空閒連接池太大,空閒連接太多,還沒使用就超時。
  3. 建議只對小報文開啓長連接。

二、location 模塊解讀

location作用:基於一個指令設置URI。

2.1 基本語法:

Syntax: location [ = | ~ | ~* | ^~ ] uri { ... }
location @name { ... }
Default:    —
Context:    server, location
  • 1
  • 2
  • 3
  • 4
  • = 精確匹配,如果找到匹配=號的內容,立即停止搜索,並立即處理請求(優先級最高)
  • ~ 區分大小寫
  • ~* 不區分大小寫
  • ^~ 只匹配字符串,不匹配正則表達式
  • @ 指定一個命名的location,一般用於內部重定義請求,location @name {…}

匹配是有優先級的,不是按照nginx的配置文件進行。

官方的例子:

location = / {
    [ configuration A ]
}
location / {
    [ configuration B ]
}
location /documents/ {
    [ configuration C ]
}
location ^~ /images/ {
    [ configuration D ]
}
location ~* \.(gif|jpg|jpeg)$ {
    [ configuration E ]
}
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15

結論:

  • / 匹配A。
  • /index.html 匹配B
  • /documents/document.html 匹配C
  • /images/1.gif 匹配D
  • /documents/1.jpg 匹配的是E。

2.1 測試用的例子:

location / {
           return 401;
        }
        location = / {
            return 402;
        }
        location /documents/ {
            return 403;
        }
        location ^~ /images/ {
            return 404;
        }
        location ~* \.(gif|jpg|jpeg)$ {
            return 500;
        }
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15

測試結果(重點看):

[root@lb01 conf]# curl -I -s -o /dev/null -w "%{http_code}\n" http://10.0.0.7/
402
[root@lb01 conf]# curl -I -s -o /dev/null -w "%{http_code}\n" http://10.0.0.7/index.html
401
[root@lb01 conf]# curl -I -s -o /dev/null -w "%{http_code}\n" http://10.0.0.7/documents/document.html 
403
[root@lb01 conf]# curl -I -s -o /dev/null -w "%{http_code}\n" http://10.0.0.7/images/1.gif
404
[root@lb01 conf]# curl -I -s -o /dev/null -w "%{http_code}\n" http://10.0.0.7/dddd/1.gif  
500
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10

結果總結:
匹配的優先順序,=>^~(匹配固定字符串,忽略正則)>完全相等>~*>>/
工作中儘量將’='放在前面

三、proxy_pass 模塊解讀

proxy_pass 指令屬於ngx_http_proxy_module 模塊,此模塊可以將請求轉發到另一臺服務器。

寫法:

proxy_pass http://localhost:8000/uri/; 
  • 1

實例一:

    upstream blog_real_servers {
         server 10.0.0.9:80  weight=5;
         server 10.0.0.10:80  weight=10;
         server 10.0.0.19:82  weight=15;
    }
    server {
       listen       80;
       server_name  blog.etiantian.org;
       location / {
        proxy_pass http://blog_real_servers;
       proxy_set_header host $host;
       }
    }
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • proxy_set_header:當後端Web服務器上也配置有多個虛擬主機時,需要用該Header來區分反向代理哪個主機名,proxy_set_header host $host;
  • proxy_set_header X-Forwarded-For :如果後端Web服務器上的程序需要獲取用戶IP,從該Header頭獲取。proxy_set_header X-Forwarded-For $remote_addr;

3.1 配置後端服務器接收前端真實IP

配置如下:

    log_format  commonlog  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';
  • 1
  • 2
  • 3

rs_apache節點的httpd.conf配置

LogFormat "\"%{X-Forwarded-For}i\" %l %u %t \"%r\" %>s %b \"%{Referer}i\" \"%{U
ser-Agent}i\"" combined修改日誌記錄
apache
LogFormat "\"%{X-Forwarded-For}i\" %l %u %t \"%r\" %>s %b" common
  • 1
  • 2
  • 3
  • 4

3.2 proxy_pass相關的優化參數

  • client_max_body_size 10m; 允許客戶端請求的最大的單個文件字節數。
  • client_body_buffer_size 128k; 緩衝區代理緩衝用戶端請求的最大字節數 可以理解爲先保存到本地再傳給用戶。
  • proxy_connect_timeout 600; 跟後端服務器連接的超時時間_發起握手等候響應超時時間。
  • proxy_read_timeout 600; 連接成功後_等候後端服務器響應時間_其實已經進入後端的排隊之中等候處理。
  • proxy_send_timeout 600; 後端服務器回傳數據時間,就是在規定時間之內後端服務器必須傳完所有的數據。
  • proxy_buffer_size 8k; 代理請求緩存區,這個緩存區間會保存用戶的頭信息以供Nginx進行規則處理,一般只要設置能保存下頭信息即可。
  • proxy_buffers 4 32k; 同上 告訴Nginx保存單個頁面使用的空間大小,假設網頁大小平均在32k以下的話。
  • proxy_busy_buffers_size 64k; 如果系統很忙的時候可以申請更大的proxy_buffers 官方推薦(proxy_buffers*2)。
  • proxy_max_temp_file_size 1024m; 當 proxy_buffers 放不下後端服務器的響應內容時,會將一部分保存到硬盤的臨時文件中,這個值用來設置最大臨時文件大小,默認1024M,它與 proxy_cache 沒有關係。大於這個值,將從upstream服務器傳回。設置爲0禁用。
  • proxy_temp_file_write_size 64k; proxy緩存臨時文件的大小 proxy_temp_path(可以在編譯的時候)指定寫到哪那個目錄。

四、健康檢查

Nginx提供了health_check語句來提供負載(upstream)時的鍵康檢查機制(注意:此語句需要設置在location上下文中)。

支持的參數有:

  • interval=time:設置兩次健康檢查之間的間隔值,默認爲5秒
  • fails=number:設置將服務器視爲不健康的連續檢查次數,默認爲1次
  • passes=number:設置一個服務器被視爲健康的連續檢查次數,默認爲1次
  • uri=uri:定義健康檢查的請求URI,默認爲”/“
  • match=name:指定匹配配置塊的名字,用記測試響應是否通過健康檢測。默認爲測試返回狀態碼爲2xx和3xx

一個簡單的設置如下,將使用默認值:

location / {
    proxy_pass http://backend;
    health_check;
}
  • 1
  • 2
  • 3
  • 4

對就應用,我們可以專門定義一個API用於健康檢查:/api/health_check,並只返回HTTP狀態碼爲200。並設置兩次檢查之間的間隔值爲1秒。這樣,health_check語句的配置如下:

health_check uri="/api/health_check" interval;
  • 1

匹配match的方法

http {
    server {
    ...
        location / {
            proxy_pass http://backend;
            health_check match=welcome;
        }
    }
match welcome {
    status 200;
    header Content-Type = text/html;
    body ~ "Welcome to nginx!";
}

}

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15

match 例子舉例

  • status 200;: status 等於 200
  • status ! 500;: status 不是 500
  • status 200 204;: status 是 200 或 204
  • status ! 301 302;: status 不是301或302。
  • status 200-399;: status 在 200 到 399之間。
  • status ! 400-599;: status 不在 400 到 599之間。
  • status 301-303 307;: status 是 301, 302, 303, 或 307。
  • header Content-Type = text/html;: “Content-Type” 得值是 text/html。
  • header Content-Type != text/html;: “Content-Type” 不是 text/html。
  • header Connection ~ close;: “Connection” 包含 close。
  • header Connection !~ close;: “Connection” 不包含 close。
  • header Host;: 請求頭包含 “Host”。
  • header ! X-Accel-Redirect;: 請求頭不包含 “X-Accel-Redirect”。
  • body ~ "Welcome to nginx!";: body 包含 “Welcome to nginx!”。
  • body !~ "Welcome to nginx!";: body 不包含 “Welcome to nginx!”。

五、一個完整的nginx實例

[root@lb01 conf]# cat nginx.conf
worker_processes  1;
events {
    worker_connections  1024;
}
http {
    include       mime.types;
    default_type  application/octet-stream;
    sendfile        on;
    keepalive_timeout  65;
    #blog lb by oldboy at 201303
    upstream blog_real_servers {
   server   10.0.0.9:80 weight=1 max_fails=1 fail_timeout=10s;
   server   10.0.0.10:80 weight=1 max_fails=2 fail_timeout=20s;
}
server {
   listen       80;
   server_name  blog.etiantian.org;
   location / {
    proxy_pass http://blog_real_servers;
    include proxy.conf;
   }
}

}
[root@lb01 conf]# cat proxy.conf
proxy_set_header Host $host;
proxy_set_header X-Forwarded-For $remote_addr;
proxy_connect_timeout 90;
proxy_send_timeout 90;
proxy_read_timeout 90;
proxy_buffer_size 4k;
proxy_buffers 4 32k;
proxy_busy_buffers_size 64k; proxy_temp_file_write_size 64k;

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34

六、擴展補充

6.1 只允許使用GET,HEAD,POST方法去請求

## Only allow these request methods ##
     if ($request_method !~ ^(GET|HEAD|POST)$ ) {
         return 444;
     }
  • 1
  • 2
  • 3
  • 4

七、實戰

7.1 根據URI及location實現動靜分離。

最終實現:

  1. /static/的URL都去訪問10.0.0.9。
  2. /dynamic/的URL都去訪問10.0.0.10。
  3. 圖片這些靜態文件去訪問10.0.0.9。
  4. /upload/的URL都去訪問10.0.0.10。
[root@lb01 conf]# cat nginx.conf
worker_processes  1;
events {
    worker_connections  1024;
}
http {
    include       mime.types;
    default_type  application/octet-stream;
    sendfile        on;
    keepalive_timeout  65;
    #blog lb by oldboy at 201303
upstream static_pools {
  server 10.0.0.9:80;
}
upstream dynamic_pools {
  server 10.0.0.10:80;
}
 upstream upload_pools {
  server 10.0.0.9:80;
}

server {
   listen       80;
   server_name  blog.biglittleant.cn;
   
    location / {
    proxy_pass http://static_pools;
    include proxy.conf;
   }

   location /static/ { 
    proxy_pass http://static_pools;
    include proxy.conf;
   }
   
    location ~* \.(gif|jpg|jpeg)$ {
     proxy_pass http://static_pools;
     include proxy.conf;
    }

   location /dynamic/ { 
    proxy_pass http://dynamic_pools;
    include proxy.conf;
   }
   location /upload/ {
    proxy_pass http://upload_pools;
    include proxy.conf;
   }
}

}

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51

7.2 實現蘋果手機和安卓手機訪問不同的地址

server {
       listen       80;
       server_name  blog.etiantian.org;
       location / {
        if ($http_user_agent ~* "android")
          {
            proxy_pass http://android_pools;
          }
        if ($http_user_agent ~* "iphone")
          {
            proxy_pass http://iphone_pools;
            }
        proxy_pass http://pc_pools;
        include extra/proxy.conf;
       }
        access_log off;
     }
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17

參考文檔

nginx-proxy_pass官網

發表評論
所有評論
還沒有人評論,想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章