在使用nginx的反向代理過程中晚吞，對(duì)于服務(wù)器節(jié)點(diǎn)的健康檢查和故障轉(zhuǎn)移很重要。
早期使用nginx的時(shí)候谋国，用來做故障轉(zhuǎn)移用到的是如下配置（比較粗暴）
upstream portals
{
server172.16.68.134:8082 max_fails=1 fail_timeout=5;
server172.16.68.135:8082 max_fails=1 fail_timeout=5;
server172.16.68.136:8082 max_fails=1 fail_timeout=5;
server172.16.68.137:8082 max_fails=1 fail_timeout=5;
}
經(jīng)過實(shí)際測試槽地，在5s內(nèi)，如果第一個(gè)服務(wù)器節(jié)點(diǎn)都不返回芦瘾，在這5s內(nèi)捌蚊，請求不會(huì)向這臺(tái)服務(wù)器轉(zhuǎn)發(fā)，5s的超時(shí)時(shí)間到了近弟，再次發(fā)起請求逢勾，就按照輪轉(zhuǎn)規(guī)則，該到這臺(tái)服務(wù)器還是會(huì)過去藐吮，這時(shí)候再經(jīng)歷5s溺拱，請求不會(huì)到這臺(tái)服務(wù)器。這樣子實(shí)際達(dá)不到想要的效果谣辞，在服務(wù)沒恢復(fù)以前迫摔，請求不到這臺(tái)服務(wù)器。
之后在網(wǎng)上找關(guān)于nginx健康檢查的模塊泥从、組件 句占，找到了淘寶的nginx_upstream_check_module。
安裝過程比較簡單躯嫉，nginx增加這個(gè)module的編譯即可纱烘。
Github地址：
https://github.com/yaoweibin/nginx_upstream_check_module
下載地址：
https://pan.baidu.com/s/1cSdTwiQPlIxGahjXzdCuEA
nginx、pcre的編譯目錄均默認(rèn)/home/soft祈餐，將下載的nginx_upstream_check_module-0.3.0.tar.gz上傳至/home/soft后解壓：
tar -zxvf nginx_upstream_check_module-0.3.0.tar.gz
重新編譯：
#進(jìn)入編譯目錄
cd /home/soft/nginx-1.10.1
#打補(bǔ)丁
patch -p0 < ../nginx_upstream_check_module-0.3.0/check_1.11.1+.patch
#確認(rèn)configure參數(shù)與現(xiàn)網(wǎng)一致擂啥，只增加一個(gè)module
./configure --prefix=/usr/local/nginx --with-pcre=/home/soft/pcre-8.36/ --with-http_stub_status_module --with-http_ssl_module --add-module=/home/soft/nginx_upstream_check_module-0.3.0/
#執(zhí)行make
make
#備份現(xiàn)網(wǎng)nginx執(zhí)行文件
cd /usr/local/nginx/sbin
mv nginx nginx.old.20181016
#拷貝升級(jí)后的執(zhí)行文件到現(xiàn)網(wǎng)目錄
cp /home/soft/nginx-1.10.1/objs/nginx /usr/local/nginx/sbin
#測試nginx版本及是否正常
./nginx -V
./nginx -t
在nginx中用到的配置是：
upstream portals {
server 192.166.62.137:8080;
server 192.166.66.85:8080;
server 192.166.62.231:8080;
server 192.166.66.88:8080;
check interval=5000 rise=2 fall=5 timeout=1000 type=http;
check_http_send"HEAD / HTTP/1.0\r\n\r\n"; check_http_expect_alive http_2xx http_3xx;
}
server {
listen 8080;
charset utf-8;
location /status {
check_status;
access_log off;
#allow 192.166.62.25;
#deny all;
}
location / {
proxy_pass http://portal_service_pool;
index index.html;
}
interval間隔5s，連續(xù)失敗5次帆阳，連續(xù)成功2次哺壶，超時(shí)時(shí)間1s，使用http協(xié)議，發(fā)送一個(gè)請求頭山宾，如果是2xx或者3xx狀態(tài)（比如200,302等）表示服務(wù)正常運(yùn)行至扰。
可以開啟注釋的配置，以使只有固定ip可以查看status頁面资锰，其他ip無法訪問此location敢课。
修改完成后reload nginx使配置生效。
sbin/nginx -s reload
觀察主動(dòng)的健康檢查效果：
在一臺(tái)服務(wù)器上執(zhí)行ab并發(fā)測試：
ab -n 20000 -c 10 "http://192.166.62.104:8080/PortalServer-App/index.html"
查看104上nginx日志:
tail -f logs/access.log
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.62.231:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.66.88:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.66.88:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 567 "-" "ApacheBench/2.3" "-" "192.166.62.137:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.62.231:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 567 "-" "ApacheBench/2.3" "-" "192.166.62.137:8080""0.000"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.66.88:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.62.231:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 567 "-" "ApacheBench/2.3" "-" "192.166.62.137:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.62.231:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.66.88:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.66.88:8080""0.002"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 567 "-" "ApacheBench/2.3" "-" "192.166.62.137:8080""0.001"
192.166.62.100 - - [16/Oct/2018:13:46:44 +0800] "GET /PortalServer-App/index.html HTTP/1.0" 200 541 "-" "ApacheBench/2.3" "-" "192.166.62.231:8080""0.001"
可見status中正常的后端可以被負(fù)載到绷杜，從而實(shí)現(xiàn)主動(dòng)健康檢查的效果直秆。
