Nginx + varnish 构建高可用CDN节点集群

jopen 10年前

一、        环境描述

Linux server A  (CentOS release 5.8 Final)  实IP:192.168.4.97  虚IP:192.168.4.96
Linux server B  (CentOS release 5.8 Final)  实IP:192.168.4.99  虚IP:192.168.4.98

域名环境(DNS轮询解析到虚IP):
server.osapub.com 192.168.4.96
server.osapub.com 192.168.4.98

二、        简单架构示意图
Nginx + varnish 构建高可用CDN节点集群
描述:前端两台NGINX做DNS轮询,通过虚IP漂移+HA监控脚本相结合,实现前端两台NGINX高可用,利用NGINX反向代理功能对后端varnish实现高可用集群。

三、软件环境搭建
3.1 编译安装nginx

?  #!/bin/bash  ####nginx 环境安装脚本,注意环境不同可能导致脚本运行出错,如果环境不同建议手工一条一条执行指令。    #创建工作目录  mkdir -p /dist/{dist,src}  cd /dist/dist    #下载安装包  wget http://bbs.osapub.com/down/google-perftools-1.8.3.tar.gz &> /dev/null  wget http://bbs.osapub.com/down/libunwind-0.99.tar.gz &> /dev/null  wget http://bbs.osapub.com/down/pcre-8.01.tar.gz &> /dev/null  wget http://bbs.osapub.com/down/nginx-1.0.5.tar.gz &> /dev/null    #------------------------------------------------------------------------  # 使用Google的开源TCMalloc库,忧化性能    cd /dist/src  tar zxf ../dist/libunwind-0.99.tar.gz  cd libunwind-0.99/    ## 注意这里不能加其它 CFLAGS加速编译参数    CFLAGS=-fPIC ./configure  make clean  make CFLAGS=-fPIC  make CFLAGS=-fPIC install  if [ "$?" == "0" ]; then          echo "libunwind-0.99安装成功." >> ./install_log.txt  else          echo "libunwind-0.99安装失败." >> ./install_log.txt          exit 1  fi    ##----------------------------------------------------------  ## 使用Google的开源TCMalloc库,提高MySQL在高并发情况下的性能    cd /dist/src  tar zxf ../dist/google-perftools-1.8.3.tar.gz  cd google-perftools-1.8.3/    CHOST="x86_64-pc-linux-gnu" CFLAGS="-march=nocona -O2 -pipe" CXXFLAGS="-march=nocona -O2 -pipe" \  ./configure    make clean  make && make install  if [ "$?" == "0" ]; then          echo "google-perftools-1.8.3安装成功." >> ./install_log.txt  else          echo "google-perftools-1.8.3安装失败." >> ./install_log.txt          exit 1  fi    echo "/usr/local/lib" > /etc/ld.so.conf.d/usr_local_lib.conf  /sbin/ldconfig    ################ 安装nginx ##########################  #安装Nginx所需的pcre库    cd /dist/src  tar zxvf ../dist/pcre-8.01.tar.gz  cd pcre-8.01/    CHOST="x86_64-pc-linux-gnu" CFLAGS="-march=nocona -O2 -pipe" CXXFLAGS="-march=nocona -O2 -pipe" \  ./configure  make && make install  if [ "$?" == "0" ]; then          echo "pcre-8.01安装成功." >> ./install_log.txt  else          echo "pcre-8.01安装失败." >> ./install_log.txt          exit 1  fi  cd ../    ## 安装Nginx  ## 为优化性能,可以安装 google 的 tcmalloc ,之前己经安装过了  ## 所以我们编译 Nginx 时,加上参数 --with-google_perftools_module  ## 然后在启动nginx前需要设置环境变量 export LD_PRELOAD=/usr/local/lib/libtcmalloc.so  ## 加上 -O2 参数也能优化一些性能  ##  ## 默认的Nginx编译选项里居然是用 debug模式的(-g参数),在 auto/cc/gcc 文件最底下,去掉那个 -g 参数  ## 就是将  CFLAGS="$CFLAGS -g"  修改为   CFLAGS="$CFLAGS"   或者直接删除这一行    cd /dist/src  rm -rf nginx-1.0.5  tar zxf ../dist/nginx-1.0.5.tar.gz  cd nginx-1.0.5/    sed -i 's#CFLAGS="$CFLAGS -g"#CFLAGS="$CFLAGS "#' auto/cc/gcc    make clean    CHOST="x86_64-pc-linux-gnu" CFLAGS="-march=nocona -O2 -pipe" CXXFLAGS="-march=nocona -O2 -pipe" \  ./configure --user=www --group=www \  --prefix=/usr/local/nginx \  --with-http_stub_status_module \  --with-google_perftools_module    make && make install  if [ "$?" == "0" ]; then          echo "nginx-1.0.5安装成功." >> ./install_log.txt  else          echo "nginx-1.0.5安装失败." >> ./install_log.txt          exit 1  fi  cd ../    #创建Nginx日志目录  mkdir -p /data/logs  chmod +w /data/logs  chown -R www:www /data/logs    cd /usr/local/nginx/  mv conf conf_bak  ln -s /data/conf/nginx/ conf    echo 'export LD_PRELOAD=/usr/local/lib/libtcmalloc.so' > /root/nginx_start  echo 'ulimit -SHn 51200' >> /root/nginx_start  echo '/usr/local/nginx/sbin/nginx -c /usr/local/nginx/conf/nginx.conf' >> /root/nginx_start    echo '/usr/local/nginx/sbin/nginx -t' > /root/nginx_reload  echo 'kill -HUP `cat /usr/local/nginx/logs/nginx.pid`' >> /root/nginx_reload    chmod 700 /root/nginx_*

3.2        编译安装varnish

#!/bin/bash    #进入工作目录  cd /dist/dist    #下载安装包  wget <a href="\"http://bbs.osapub.com/down/varnish-3.0.0.tar.gz\"" target="\"_blank\"">http://bbs.osapub.com/down/varnish-3.0.0.tar.gz</a> &> /dev/null  cd /dist/src  rm -fr varnish-3.0.0  export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig  tar zxvf ../dist/varnish-3.0.0.tar.gz  cd varnish-3.0.0  #编译参数可以根据自己需要定制  ./configure -prefix=/usr/local/varnish -enable-debugging-symbols -enable-developer-warnings -enable-dependency-tracking  make && make install  if [ "$?" == "0" ]; then          echo "varnish-3.0.0安装成功." >> ./install_log.txt  else          echo "varnish-3.0.0安装失败." >> ./install_log.txt          exit 1  fi    #设置启动、重启脚本  cat > /root/varnish_restart.sh <<EOF  #!/bin/bash  pkill varnish  pkill varnish  ulimit -SHn 51200  /usr/local/varnish/sbin/varnishd -n /data/varnish -f /usr/local/varnish/vcl.conf -a 0.0.0.0:81 -s malloc,12G -g www -u www -w 4000,12000,10 -T 127.0.0.1:3500 -p http_max_hdr=256 -p http_req_hdr_len=8192  EOF    cat > /root/varnish_start.sh <<EOF  #!/bin/bash  ulimit -SHn 51200  /usr/local/varnish/sbin/varnishd -n /data/varnish -f /usr/local/varnish/vcl.conf -a 0.0.0.0:81 -s malloc,12G -g www -u www -w 4000,12000,10 -T 127.0.0.1:3500 -p http_max_hdr=256 -p http_req_hdr_len=8192  EOF    chmod 755 /root/varnish*


三、        配置varnish一些参数说明:
Backend servers
Varnish有后端(或称为源)服务器的概念。后端服务器是指Varnish提供加速服务的那台,通常提供内容。
第一件要做的事情是告诉Varnish,哪里能找到要加速的内容。
vcl_recv
vcl_recv是在请求开始时调用的。完成该子程序后,请求就被接收并解析了。用于确定是否需要服务请求,怎么服务,如果可用,使用哪个后端。
在vcl_recv中,你也可以修改请求。通常你可以修改cookie,或添加/移除请求头信息。
注意在vcl_recv中,只可以使用请求对象req。

vcl_fetch
vcl_fetch是在文档从后端被成功接收后调用的。通常用于调整响应头信息,触发ESI处理,万一请求失败就换个后端服务器。
在vcl_fecth中,你还可以使用请求对象req。还有个后端响应对象beresp。Beresp包含了后端的HTTP头信息。

varnish 3.X 配置参考文档:http://anykoro.sinaapp.com/?p=261

编辑:/usr/local/varnish/vcl.conf ,文件不存在则创建。注意:Server A 与server B 配置一致!

配置详情如下:

?  ###########后台代理服务器########    backend server_osapub_com{          .host = "192.168.4.97";          .port = "82";  }    acl purge {          "localhost";                  "127.0.0.1";  }    #############################################    sub vcl_recv {         #################BAN##########################begin           if (req.request == "BAN") {                  # Same ACL check as above:                  if (!client.ip ~ purge) {                          error 405 "Not allowed.";                  }                  ban("req.http.host == " + req.http.host +                        "&& req.url == " + req.url);                  error 200 "Ban added";          }       #################BAN##########################end             ###############配置域名##################################            if(req.http.host ~ "^server.osapub.com" ||req.http.host ~ ".osapub.com" ) {                    #使用哪一组后台服务器                          set req.backend = server_osapub_com;                    }                    ##################################################            if (req.restarts == 0) {          if (req.http.x-forwarded-for) {             set req.http.X-Forwarded-For = req.http.X-Forwarded-For + ", " + client.ip;           }           else {              set req.http.X-Forwarded-For = client.ip;           }       }            if (req.http.Accept-Encoding) {          if (req.url ~ "\.(jpg|jpeg|png|gif|gz|tgz|bz2|tbz|mp3|ogg|swf)$") {              remove req.http.Accept-Encoding;          } elsif (req.http.Accept-Encoding ~ "gzip") {              set req.http.Accept-Encoding = "gzip";}            elsif (req.http.Accept-Encoding ~ "deflate") {              set req.http.Accept-Encoding = "deflate";}          else {           remove req.http.Accept-Encoding;        }      }       if (req.http.Cache-Control ~ "no-cache") {           return (pass);       }       if (req.request != "GET" &&         req.request != "HEAD" &&         req.request != "PUT" &&         req.request != "POST" &&         req.request != "TRACE" &&         req.request != "OPTIONS" &&         req.request != "DELETE") {           return (pipe);       }         if (req.request != "GET" && req.request != "HEAD") {                          return(pass);                  }                            if (req.http.Authorization || req.http.Cookie ||req.http.Authenticate) {                   return (pass);       }                     if (req.request == "GET" && req.url ~ "(?i)\.php($|\?)"){                  return (pass);               }            if (req.url ~ "\.(css|jpg|jpeg|png|gif|gz|tgz|bz2|tbz|mp3|ogg|swf)$") {                  unset req.http.Cookie;          }                        return (lookup);    }    sub vcl_fetch {            set beresp.grace = 5m;            if (beresp.status == 404 || beresp.status == 503 || beresp.status == 500 || beresp.status == 502) {                  set beresp.http.X-Cacheable = "NO: beresp.status";                  set beresp.http.X-Cacheable-status = beresp.status;                  return (hit_for_pass);          }            #决定哪些头不缓存        if (req.url ~ "\.(php|shtml|asp|aspx|jsp|js|ashx)$") {                  return (hit_for_pass);           }            if (req.url ~ "\.(jpg|jpeg|png|gif|gz|tgz|bz2|tbz|mp3|ogg|swf)$") {                  unset beresp.http.set-cookie;          }            if (beresp.ttl <= 0s) {                  set beresp.http.X-Cacheable = "NO: !beresp.cacheable";                  return (hit_for_pass);          }          else {                  unset beresp.http.expires;          }            return (deliver);    }    sub vcl_deliver {    if (resp.http.magicmarker) {         unset resp.http.magicmarker;         set resp.http.age = "0";   }    # add cache hit data  if (obj.hits > 0) {       set resp.http.X-Cache = "HIT";     set resp.http.X-Cache-Hits = obj.hits;  }  else {       set resp.http.X-Cache = "MISS";  }    # hidden some sensitive http header returning to client, when the cache server received from backend server response  remove resp.http.X-Varnish;  remove resp.http.Via;  remove resp.http.Age;  remove resp.http.X-Powered-By;  remove resp.http.X-Drupal-Cache;    return (deliver);  }    sub vcl_error {     if (obj.status == 503 && req.restarts < 5) {     set obj.http.X-Restarts = req.restarts;     return (restart);   }    }    sub vcl_hit {    if (req.http.Cache-Control ~ "no-cache") {      if (! (req.http.Via || req.http.User-Agent ~ "bot|MSIE")) {        set obj.ttl = 0s;        return (restart);     }  }    return(deliver);  }

配置完成后执行:/root/varnish_start.sh 如果启动成功则表示配置无误!
Nginx + varnish 构建高可用CDN节点集群

四、        配置nginx编辑: /usr/local/nginx/conf/nginx.conf ,nginx 的配置方法请参考网上文档资料。
Server A(192.168.4.97) nginx配置如下:

?  user  www www;  worker_processes 16;  error_log /logs/nginx/nginx_error.log crit;    #Specifies the value for maximum file descriptors that can be opened by this process.  worker_rlimit_nofile 65500;    events  {          use epoll;            worker_connections 65500;  }    http  {          include       mime.types;          default_type  application/octet-stream;            server_names_hash_bucket_size 128;          client_header_buffer_size 64k;          large_client_header_buffers 4 64k;          client_max_body_size 10m;            server_tokens off;            expires       1h;            sendfile on;          tcp_nopush     on;          keepalive_timeout 60;          tcp_nodelay on;        fastcgi_connect_timeout 200;      fastcgi_send_timeout 300;      fastcgi_read_timeout 600;      fastcgi_buffer_size 128k;      fastcgi_buffers 4 64k;      fastcgi_busy_buffers_size 128k;      fastcgi_temp_file_write_size 128k;      fastcgi_temp_path /dev/shm;          gzip on;          gzip_min_length  2048;          gzip_buffers     4 16k;          gzip_http_version 1.1;          gzip_types  text/plain  text/css  application/xml application/x-javascript ;            log_format  access  '$remote_addr - $remote_user [$time_local] "$request" '                            '$status $body_bytes_sent "$http_referer" '                            '"$http_user_agent" $http_x_forwarded_for';    #################  include  ###################             upstream varnish_server {                  server 127.0.0.1:81 weight=2 max_fails=3 fail_timeout=30s;                      server 192.168.4.99:81 weight=2 max_fails=3 fail_timeout=30s;                  ip_hash;          }                    server          {                  listen       80;                  server_name  192.168.4.96 192.168.4.98;                               index index.html index.htm index.php index.aspx;                  location /                  {                          proxy_set_header   Host             $host;                          proxy_set_header   X-Real-IP        $remote_addr;                          proxy_set_header   X-Forwarded-For  $proxy_add_x_forwarded_for;                            proxy_connect_timeout 200;                          proxy_send_timeout 300;                          proxy_read_timeout 500;                          proxy_buffer_size 256k;                          proxy_buffers 4 128k;                          proxy_busy_buffers_size 256k;                          proxy_temp_file_write_size 256k;                          proxy_temp_path  /dev/shm;                            proxy_pass  http://varnish_server;                            expires off;                            access_log  /logs/nginx/192.168.4.96.log  access;                  }          }                    server          {                listen       82;                server_name  192.168.4.97;                index index.html index.htm index.php;                root  /data/web/awstats/www;                  location ~ .*\.php$                {                     include fcgi.conf;                     fastcgi_pass  127.0.0.1:10080;                     fastcgi_index index.php;                     expires off;                }                access_log  /logs/nginx/awstats.osapub.com.log  access;          }                   }

server B (192.168.4.99)配置如下:

?  user  www www;  worker_processes 16;  error_log /logs/nginx/nginx_error.log crit;    #Specifies the value for maximum file descriptors that can be opened by this process.  worker_rlimit_nofile 65500;    events  {          use epoll;            worker_connections 65500;  }    http  {          include       mime.types;          default_type  application/octet-stream;            server_names_hash_bucket_size 128;          client_header_buffer_size 64k;          large_client_header_buffers 4 64k;          client_max_body_size 10m;            server_tokens off;            expires       1h;            sendfile on;          tcp_nopush     on;          keepalive_timeout 60;          tcp_nodelay on;        fastcgi_connect_timeout 200;      fastcgi_send_timeout 300;      fastcgi_read_timeout 600;      fastcgi_buffer_size 128k;      fastcgi_buffers 4 64k;      fastcgi_busy_buffers_size 128k;      fastcgi_temp_file_write_size 128k;      fastcgi_temp_path /dev/shm;          gzip on;          gzip_min_length  2048;          gzip_buffers     4 16k;          gzip_http_version 1.1;          gzip_types  text/plain  text/css  application/xml application/x-javascript ;            log_format  access  '$remote_addr - $remote_user [$time_local] "$request" '                            '$status $body_bytes_sent "$http_referer" '                            '"$http_user_agent" $http_x_forwarded_for';    #################  include  ###################             upstream varnish_server {                  server 127.0.0.1:81 weight=2 max_fails=3 fail_timeout=30s;                      server 192.168.4.97:81 weight=2 max_fails=3 fail_timeout=30s;                  ip_hash;          }                    server          {                  listen       80;                  server_name  192.168.4.96 192.168.4.98;                               index index.html index.htm index.php index.aspx;                  location /                  {                          proxy_set_header   Host             $host;                          proxy_set_header   X-Real-IP        $remote_addr;                          proxy_set_header   X-Forwarded-For  $proxy_add_x_forwarded_for;                            proxy_connect_timeout 200;                          proxy_send_timeout 300;                          proxy_read_timeout 500;                          proxy_buffer_size 256k;                          proxy_buffers 4 128k;                          proxy_busy_buffers_size 256k;                          proxy_temp_file_write_size 256k;                          proxy_temp_path  /dev/shm;                            proxy_pass  http://varnish_server;                            expires off;                            access_log  /logs/nginx/192.168.4.98.log  access;                  }          }                    server          {                listen       82;                server_name  192.168.4.99;                index index.html index.htm index.php;                root  /data/web/awstats/www;                  location ~ .*\.php$                {                     include fcgi.conf;                     fastcgi_pass  127.0.0.1:10080;                     fastcgi_index index.php;                     expires off;                }                access_log  /logs/nginx/awstats.osapub.com.log  access;          }                   }

启动nginx:
/root/nginx_start
启动正常会监听80,82端口!
五、        运行nginx ha脚本

server A 的脚本下载地址:http://bbs.osapub.com/down/server_a_ha.tar.gz
解压力后得到三个脚本:
nginx_watchdog.sh
nginxha.sh
nginx_ha1.sh

server B 的脚本下载地址:http://bbs.osapub.com/down/server_b_ha.tar.gz
解压力后得到三个脚本:
nginx_watchdog.sh
nginxha.sh
nginx_ha2.sh

注意:脚本建议放到:/data/sh/ha 目录下,否则需要修改nginxha.sh  里面的程序路径!

六、        测试1,测试之前先把nginx 负载去掉,可以在前面加#号暂时注释,只保留本机解析,确保测试结果准确。
Nginx + varnish 构建高可用CDN节点集群

2,修改本机host文件
添加:192.168.4.99 server.osapub.com 到末尾,保存后访问。
Nginx + varnish 构建高可用CDN节点集群

3,正常结果显示如下:
Nginx + varnish 构建高可用CDN节点集群

到这里整个架构基本能运行起来了,根据大家的实际需求,对配置文件进行调优,HA脚本也可以进一步调优,关于报警,请参考社区自动安装mutt报警的脚本!

来自:http://my.oschina.net/imot/blog/289208