LVS+Heartbeat 高可用集群方案操作记录

Heartbeat 项目是 Linux-HA 工程的一个组成部分，它实现了一个高可用集群系统。心跳服务和集群通信是高可用集群的两个关键组件，在 Heartbeat 项目里，由 heartbeat 模块实现了这两个功能。

Heartbeat的高可用集群采用的通信方式是udp协议和串口通信，而且heartbeat插件技术实现了集群间的串口、多播、广播和组播通信。它实现了HA 功能中的核心功能——心跳，将Heartbeat软件同时安装在两台服务器上，用于监视系统的状态，协调主从服务器的工作，维护系统的可用性。它能侦测服务器应用级系统软件、硬件发生的故障，及时地进行错误隔绝、恢复；通过系统监控、服务监控、IP自动迁移等技术实现在整个应用中无单点故障，简单、经济地确保重要的服务持续高可用性。 Heartbeat采用虚拟IP地址映射技术实现主从服务器的切换对客户端透明的功能。但是单一的heartbeat是无法提供健壮的服务的，所以这里结合使用lvs进行负载均衡。

LVS是Linux Virtual Server的简写, 意即Linux虚拟服务器，是一个虚拟的服务器集群系统。说到lvs就得提到ipvs (ipvsadm命令)，ipvs 是 lvs集群系统的核心软件,它的主要作用是安装在 Load Balancer 上,把发往 Virtual IP 的请求转发到 Real Server 上。

ldirectord是配合lvs作为一种健康检测机制，要不负载均衡器在节点挂掉后依然没有检测的功能。

案例架构草图如下:

1) 基本环境准备 (centos6.9系统)

172.16.60.206(eth0)    HA主节点(ha-master)       heartbeat, ipvsadm, ldirectord
172.16.60.207(eth0)    HA备节点(ha-slave)        heartbeat, ipvsadm, ldirectord
172.16.60.111          VIP地址
172.16.60.204(eth0)    后端节点1(rs-204)         nginx, realserver
172.16.60.205(eth0)    后端节点2(rs-205)         nginx, realserver
 
1) 关闭防火墙和selinux (四台节点机都操作)
[root@ha-master ~]# /etc/init.d/iptables stop
[root@ha-master ~]# setenforce 0
[root@ha-master ~]# vim /etc/sysconfig/selinux
SELINUX=disabled
 
2) 设置主机名和绑定hosts (两台HA节点机器都操作)
主节点操作
[root@ha-master ~]# hostname ha-master
[root@ha-master ~]# vim /etc/sysconfig/network
HOSTNAME=ha-master
[root@ha-master ~]# vim /etc/hosts
172.16.60.206 ha-master
172.16.60.207 ha-slave
 
备节点操作
[root@ha-slave ~]# hostname ha-slave
[root@ha-slave ~]# vim /etc/sysconfig/network
HOSTNAME=ha-slave
[root@ha-slave ~]# vim /etc/hosts
172.16.60.206 ha-master
172.16.60.207 ha-slave
 
3) 设置ip路由转发功能 (四台节点机器都设置)
[root@ha-master ~]# echo 1 > /proc/sys/net/ipv4/ip_forward
[root@ha-master ~]# vim /etc/sysctl.conf
net.ipv4.ip_forward = 1
[root@ha-master ~]# sysctl -p

2) 安装配置 Heartbeat (两台HA节点机都操作)

1) 首先安装heartbeat (HA主备两个节点都要同样操作)
分别下载epel-release-latest-6.noarch.rpm 和 ldirectord-3.9.5-3.1.x86_64.rpm
下载地址:　https://pan.baidu.com/s/1IvCDEFLCBYddalV89YvonQ
提取密码:　gz53
   
[root@ha-master ~]# ll epel-release-latest-6.noarch.rpm
-rw-rw-r-- 1 root root 14540 Nov  5  2012 epel-release-latest-6.noarch.rpm
[root@ha-master ~]# ll ldirectord-3.9.5-3.1.x86_64.rpm
-rw-rw-r-- 1 root root 90140 Dec 24 15:54 ldirectord-3.9.5-3.1.x86_64.rpm
   
[root@ha-master ~]# yum install -y epel-release
[root@ha-master ~]# rpm -ivh epel-release-latest-6.noarch.rpm --force
[root@ha-master ~]# yum install -y heartbeat* libnet
[root@ha-master ~]# yum install -y ldirectord-3.9.5-3.1.x86_64.rpm      ＃因为依赖比较多，　所以直接采用yum方式安装
   
2) 配置heartbeat (HA主备两个节点都要操作)
安装完heartbeat后系统会生成一个/etc/ha.d/目录，此目录用于存放heartbeat的有关配置文件。
Heartbeat自带配置文件的注释信息较多，在此手工编写有关配置文件，heartbeat常用配置文件有四个，分别是：
ha.cf：heartbeat主配置文件
ldirectord.cf：资源管理文件
haresources：本地资源文件
authkeys：认证文件
   
[root@ha-master ~]# cd /usr/share/doc/heartbeat-3.0.4/
[root@ha-master heartbeat-3.0.4]# cp authkeys ha.cf haresources /etc/ha.d/
   
[root@ha-master heartbeat-3.0.4]# cd /usr/share/doc/ldirectord-3.9.5
[root@ha-master ldirectord-3.9.5]# cp ldirectord.cf /etc/ha.d/
[root@ha-master ldirectord-3.9.5]# cd /etc/ha.d/
[root@ha-master ha.d]# ll
total 56
-rw-r--r-- 1 root root   645 Dec 24 21:37 authkeys
-rw-r--r-- 1 root root 10502 Dec 24 21:37 ha.cf
-rwxr-xr-x 1 root root   745 Dec  3  2013 harc
-rw-r--r-- 1 root root  5905 Dec 24 21:37 haresources
-rw-r--r-- 1 root root  8301 Dec 24 21:38 ldirectord.cf
drwxr-xr-x 2 root root  4096 Dec 24 21:28 rc.d
-rw-r--r-- 1 root root   692 Dec  3  2013 README.config
drwxr-xr-x 2 root root  4096 Dec 24 21:28 resource.d
-rw-r--r-- 1 root root  2082 Mar 24  2017 shellfuncs
   
3) 配置heartbeat的主配置文件ha.cf  （HA主备节点配置一样）
[root@ha-master ha.d]# pwd
/etc/ha.d
[root@ha-master ha.d]# cp ha.cf ha.cf.bak
[root@ha-master ha.d]# > ha.cf
[root@ha-master ha.d]# vim ha.cf
debugfile /var/log/ha-debug
logfile /var/log/ha-log         #日志存放位置
#crm yes                            #是否开启集群资源管理功能
logfacility        local0         #记录日志等级
keepalive 2                         #心跳的时间间隔，默认时间单位为秒
deadtime 5                         #超出该时间间隔未收到对方节点的心跳，则认为对方已经死亡。
warntime 3                         #超出该时间间隔未收到对方节点的心跳，则发出警告并记录到日志中，但此时不会切换
initdead 10          #在某些系统上，系统启动或重启之后需要经过一段时间网络才能正常工作，该选项用于解决这种情况产生的时间间隔。取值至少为deadtime的两倍。
udpport  694        #设置广播通信使用的端口，694为默认使用的端口号。
bcast        eth0               # Linux指定心跳使用以太网广播方式，并在eth0上进行广播。"#"后的要完全删除，要不然要出错。
ucast eth0 172.16.60.207       #采用网卡eth0的UDP多播来组织心跳，后面跟的IP地址应该为双机中对方的IP地址!!!!!
auto_failback on            #在该选项设为on的情况下，一旦主节点恢复运行，则自动获取资源并取代备用节点。off主节点恢复后变为备用节点，备用为主节点!!!!!
#stonith_host *     baytech 10.0.0.3 mylogin mysecretpassword
#stonith_host ken3  rps10 /dev/ttyS1 kathy 0
#stonith_host kathy rps10 /dev/ttyS1 ken3 0
#watchdog /dev/watchdog        
node   ha-master           #主机节点名，可通过"uname -n"查看，默认为主节点!!!!!
node   ha-slave              #备用机节点名，默认为次节点，要注意顺序!!!!
#ping 172.16.60.207         # 选择ping节点，选择固定路由作为节点。ping节点仅用来测试网络连接。一般选择这行ping测试就行, 下面一行注释掉.
ping_group group1 172.16.60.204 172.16.60.205     #这个地址并不是双机中的两个节点地址，而是仅仅用来测试网络的连通性. 当这两个IP 都不能ping通时，对方即开始接管资源。
respawn root /usr/lib64/heartbeat/ipfail                    #选配项。其中rootr表示启动ipfail进程的身份。要确保/usr/lib64/heartbeat/ipfail这个路径正确(可以用find命令搜索出来), 否则heartbeat启动失败
apiauth ipfail gid=root uid=root
 
============================温馨提示================================
HA备节点的ha.cf文件只需要将上面配置中的ucast一行内容改为"ucast eth0 172.16.60.206" 即可, 其他配置内容和上面HA主节点的ha.cf完全一样!
 
4) 配置heartbeat的认证文件authkeys （HA主备节点配置必须一致）
[root@ha-master ~]# cd /etc/ha.d/
[root@ha-master ha.d]# cp authkeys authkeys.bak
[root@ha-master ha.d]# >authkeys
auth 3                                                      #auth后面指定的数字，下一行必须作为关键字再次出现! 一共有"1", "2","3" 三行, 这里选择"3"关键字, 选择"1"和"2"关键字也行, HA主备节点必须一致!
#1 crc
#2 sha1 HI!
3 md5 Hello!
   
必须将该文件授权为600
[root@ha-master ha.d]# chmod 600 authkeys
[root@ha-master ha.d]# ll authkeys
-rw------- 1 root root 20 Dec 25 00:16 authkeys
 
5) 修改heartbeat的资源文件haresources (HA主备节点配置必须完全一致)
[root@ha-slave ha.d]# cp haresources haresources.bak
[root@ha-slave ha.d]# >haresources
[root@ha-slave ha.d]# vim haresources          # 在文件结尾添加下面一行内容. 由于该文件默认全是注释,可以先清空该文件, 然后添加下面这一行内容
ha-master IPaddr::172.16.60.111 ipvsadm ldirectord     
 
配置说明:
上面设置ha-maser为主节点, 集群VIP为172.16.60.111, ipvsadm ldirectord为所指定需要监视的应用服务.
这样启动heartbeat服务的时候, 会自动启动ipvsadm和ldirectord服务.
ipvsadm服务的配置文件为/etc/sysconfig/ipvsadm, 后面会配置这个.
ldirectord 服务的配置文件为/etc/ha.d/ldirectord.cf, 后面会配置这个
 
6) 配置heartbeat的监控文件ldirectord.cf (HA主备节点配置必须完全一致)
ldirectord，用于监控在lvs集群的真实服务。ldirectord是和heartbeat相结合的一个服务，可以作为heartbeat的一个启动服务。
Ldirectord 的作用是监测 Real Server，当 Real Server失效时，把它从 Load Balancer列表中删除，恢复时重新添加。
将ldrectord的配置文件复制到/etc/ha.d下，因为默认没有放到这个路径下, 并且在ldirectord.cf文件中要配置"quiescent=no" 。
  
[root@ha-master ha.d]# cp ldirectord.cf ldirectord.cf.bak
[root@ha-master ha.d]# vim ldirectord.cf
checktimeout=3      #判定realserver出错时间
checkinterval=1      #指定ldirectord在两次检查之间的间隔时间,即主从切换的时间间隔
autoreload=yes       #是否自动重载配置文件
logfile="/var/log/ldirectord.log"     #指定ldirectord的日志文件路径
#logfile="local0"
#emailalert="root@30920.cn"
#emailalertfreq=3600
#emailalertstatus=all
quiescent=no        #如果一个realserver节点在checktimeout设置的时间周期内没响应，将会被踢除，中断现有客户端的连接。 设置为yes, 则出问题的realserver节点不会被踢出, 只是新的连接不能到达。
 
virtual=172.16.60.111:80     #指定虚拟IP，注意在virtual这行后面的行必须缩进一个tab字符进行标记!! 否则极有可能因为格式配置不正确而导致ldirectord启动失败
        real=172.16.60.204:80 gate   #gate为lvs的DR模式，ipip表示TUNL模式，masq表示NAT模式
        real=172.16.60.205:80 gate   #当所有RS机器不能访问的时候WEB重写向地址; 即表示realserver全部失败，vip指向本机80端口
        fallback=127.0.0.1:80 gate     #指定服务类型，这里对HTTP进行负载均衡
        service=http         #指定服务类型，这里对HTTP进行负载均衡
        scheduler=wlc      #指定调度算法，这里的算法一定要和lvs脚本(/etc/sysconfig/ipvsadm)的算法一样
        persistent=600     #持久链接：表示600s之内同一个客户端ip将访问同一台realserver. 除非这个realserver出现故障,才会将请求转发到另一个realserver
        #netmask=255.255.255.255
        protocol=tcp          # 指定协议
        checktype=negotiate   #指定检查类型为协商 (或者执行检查类型为negotiate, 表示通过交互来判断服务是否正常)
        checkport=80        # 监控的端口
        request="lvs_testpage.html"   #请求监控地址, 这个文件一定要放到后端realserver监控端口的根目录下, 即放到两台realserver的nginx根目录下 
        receive="Test HA Page"      #指定请求和应答字符串，也就是上面lvs_testpage.html的内容
        #virtualhost=www.x.y.z       #虚拟服务器的名称可任意指定
 
============================温馨提示======================================
配置如上，通过virtual来定义vip，接下来是定义real service的节点，fallback是当所有real挂掉后，访问请求到本机的80端口上去，一般这个页面显示服务器正在维护等界面。
service表示；调度的服务，scheduler是调度算法，protocol是定义协议，checktype是检查类型为协商，checkport就是检查的端口，也就是健康检查。
  
上面在/etc/ha.d/ldirectord.cf文件里定义了一个80端口的代理转发, 如果还有其他端口, 比如3306,
 则只需要在下面再添加一个"virtual=172.16.60.111:3306 ...."类似上面的配置即可! 配置案例在备份的ldirectord.cf.bak文件里有.
 
ldirectord.cf文件的配置, 最好按照这个文件里的配置范例去修改, 不要全部清空后自行添加, 否则容易因为配置格式问题导致ldirectord服务启动失败!
 
使用status查看ldirectord服务, 只要不出现报错信息, 就说明ldirectord.cf文件配置没有问题了!
[root@ha-master ha.d]# /etc/init.d/ldirectord status

3) 安装配置 LVS (两台HA节点机操作一致)

1) 安装lvs依赖
[root@ha-master ~]# yum install -y libnl* popt*
   
查看是否加载lvs模块
[root@ha-master ~]# modprobe -l |grep ipvs
kernel/net/netfilter/ipvs/ip_vs.ko
kernel/net/netfilter/ipvs/ip_vs_rr.ko
kernel/net/netfilter/ipvs/ip_vs_wrr.ko
kernel/net/netfilter/ipvs/ip_vs_lc.ko
kernel/net/netfilter/ipvs/ip_vs_wlc.ko
kernel/net/netfilter/ipvs/ip_vs_lblc.ko
kernel/net/netfilter/ipvs/ip_vs_lblcr.ko
kernel/net/netfilter/ipvs/ip_vs_dh.ko
kernel/net/netfilter/ipvs/ip_vs_sh.ko
kernel/net/netfilter/ipvs/ip_vs_sed.ko
kernel/net/netfilter/ipvs/ip_vs_nq.ko
kernel/net/netfilter/ipvs/ip_vs_ftp.ko
kernel/net/netfilter/ipvs/ip_vs_pe_sip.ko
   
2) 下载并安装LVS
[root@ha-master ~]# cd /usr/local/src/
[root@ha-master src]# unlink /usr/src/linux
[root@ha-master src]# ln -s /usr/src/kernels/2.6.32-431.5.1.el6.x86_64/ /usr/src/linux
[root@ha-master src]# wget http://www.linuxvirtualserver.org/software/kernel-2.6/ipvsadm-1.26.tar.gz
[root@ha-master src]# tar -zvxf ipvsadm-1.26.tar.gz
[root@ha-master src]# cd ipvsadm-1.26
[root@ha-master ipvsadm-1.26]# make && make install
   
LVS安装完成，查看当前LVS集群
[root@ha-master ipvsadm-1.26]# ipvsadm -L -n
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
 
3) 添加lvs的管理脚本(ipvsadm)
ipvsadm服务的配置文件是/etc/sysconfig/ipvsadm
[root@ha-master ha.d]# vim /etc/sysconfig/ipvsadm
#!/bin/bash
# description: start LVS of DirectorServer
#Written by :NetSeek http://www.linuxtone.org
GW=172.16.60.1                                   #这个是VIP所在网段的网段地址
   
# website director vip.
WEB_VIP=172.16.60.111  
WEB_RIP1=172.16.60.204
WEB_RIP2=172.16.60.205
  
   
. /etc/rc.d/init.d/functions
   
logger $0 called with $1
   
case "$1" in
   
start)
        # Clear all iptables rules.
         /sbin/iptables -F
        # Reset iptables counters.
         /sbin/iptables -Z
         # Clear all ipvsadm rules/services.
         /sbin/ipvsadm -C
   
 #set lvs vip for dr
        /sbin/ipvsadm --set 30 5 60
        /sbin/ifconfig eth0:0 $WEB_VIP broadcast $WEB_VIP netmask 255.255.255.255 up
        /sbin/route add -host $WEB_VIP dev eth0:0
                /sbin/ipvsadm -A -t $WEB_VIP:80 -s wlc -p 600
                /sbin/ipvsadm -a -t $WEB_VIP:80 -r $WEB_RIP1:80 -g
                /sbin/ipvsadm -a -t $WEB_VIP:80 -r $WEB_RIP2:80 -g
 
        touch /var/lock/subsys/ipvsadm >/dev/null 2>&1
          
        # set Arp
                /sbin/arping -I eth0 -c 5 -s $WEB_VIP $GW >/dev/null 2>&1  
       ;;
stop)
        /sbin/ipvsadm -C
        /sbin/ipvsadm -Z
        ifconfig eth0:0 down
        route del $WEB_VIP  >/dev/null 2>&1
        rm -rf /var/lock/subsys/ipvsadm >/dev/null 2>&1
                /sbin/arping -I eth0 -c 5 -s $WEB_VIP $GW
        echo "ipvsadm stoped"
       ;;
   
status)
   
        if [ ! -e /var/lock/subsys/ipvsadm ];then
                echo "ipvsadm is stoped"
                exit 1
        else
                ipvsadm -ln
                echo "..........ipvsadm is OK."
        fi
      ;;
   
*)
        echo "Usage: $0 {start|stop|status}"
        exit 1
esac
   
exit 0
 
 
===============温馨提示=================
上面配置中的"-p 600"的意思是会话保持时间为600秒，这个应该和ldirectord.cf文件配置一致 (还有lvs策略也要一致, 如这里的lwc)
 
授权脚本执行权限
[root@ha-master ha.d]# chmod 755 /etc/sysconfig/ipvsadm

4) realserver 节点配置

1) 在realserver节点上编写LVS启动脚本 (两个realserver节点操作完全一致)
[root@rs-204 ~]# vim /etc/init.d/realserver
#!/bin/sh
VIP=172.16.60.111    
. /etc/rc.d/init.d/functions
       
case "$1" in
# 禁用本地的ARP请求、绑定本地回环地址
start)
    /sbin/ifconfig lo down
    /sbin/ifconfig lo up
    echo "1" >/proc/sys/net/ipv4/conf/lo/arp_ignore
    echo "2" >/proc/sys/net/ipv4/conf/lo/arp_announce
    echo "1" >/proc/sys/net/ipv4/conf/all/arp_ignore
    echo "2" >/proc/sys/net/ipv4/conf/all/arp_announce
    /sbin/sysctl -p >/dev/null 2>&1
    /sbin/ifconfig lo:0 $VIP netmask 255.255.255.255 up  
    /sbin/route add -host $VIP dev lo:0
    echo "LVS-DR real server starts successfully.\n"
    ;;
stop)
    /sbin/ifconfig lo:0 down
    /sbin/route del $VIP >/dev/null 2>&1
    echo "1" >/proc/sys/net/ipv4/conf/lo/arp_ignore
    echo "2" >/proc/sys/net/ipv4/conf/lo/arp_announce
    echo "1" >/proc/sys/net/ipv4/conf/all/arp_ignore
    echo "2" >/proc/sys/net/ipv4/conf/all/arp_announce
echo "LVS-DR real server stopped.\n"
    ;;
status)
    isLoOn=`/sbin/ifconfig lo:0 | grep "$VIP"`
    isRoOn=`/bin/netstat -rn | grep "$VIP"`
    if [ "$isLoON" == "" -a "$isRoOn" == "" ]; then
        echo "LVS-DR real server has run yet."
    else
        echo "LVS-DR real server is running."
    fi
    exit 3
    ;;
*)
    echo "Usage: $0 {start|stop|status}"
    exit 1
esac
exit 0
   
   
启动两台realserver节点的realserver脚本
[root@rs-204 ~]# chmod 755 /etc/init.d/realserver
[root@rs-204 ~]# ll /etc/init.d/realserver
-rwxr-xr-x 1 root root 1278 Dec 24 13:40 /etc/init.d/realserver
   
[root@rs-204 ~]# /etc/init.d/realserver start
LVS-DR real server starts successfully.\n
   
设置开机启动
[root@rs-204 ~]# echo "/etc/init.d/realserver" >> /etc/rc.local
   
查看, 发现两台realserver节点上的lo:0上已经配置了vip地址
[root@rs-204 ~]# ifconfig
...........
lo:0      Link encap:Local Loopback
          inet addr:172.16.60.111  Mask:255.255.255.255
          UP LOOPBACK RUNNING  MTU:65536  Metric:1
   
   
2) 接着部署两台realserver的web测试环境  (两个realserver节点安装操作一致)
采用yum方式安装nginx (先安装nginx的yum源)
[root@rs-204 ~]# rpm -ivh http://nginx.org/packages/centos/6/noarch/RPMS/nginx-release-centos-6-0.el6.ngx.noarch.rpm
[root@rs-204 ~]# yum install nginx
   
realserver01的nginx配置
[root@rs-204 ~]# cd /etc/nginx/conf.d/
[root@rs-204 conf.d]# cat default.conf
[root@rs-204 conf.d]# >/usr/share/nginx/html/index.html
[root@rs-204 conf.d]# vim /usr/share/nginx/html/index.html
this is test page of realserver01:172.16.60.204
   
[root@rs-204 conf.d]# vim /usr/share/nginx/html/lvs_testpage.html
Test HA Page
   
[root@rs-204 conf.d]# /etc/init.d/nginx start
Starting nginx:                                            [  OK  ]
[root@rs-204 conf.d]# lsof -i:80
COMMAND   PID  USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
nginx   31944  root    6u  IPv4  91208      0t0  TCP *:http (LISTEN)
nginx   31945 nginx    6u  IPv4  91208      0t0  TCP *:http (LISTEN)
   
realserver02的nginx配置
[root@rs-205 src]# cd /etc/nginx/conf.d/
[root@rs-205 conf.d]# cat default.conf
[root@rs-205 conf.d]# >/usr/share/nginx/html/index.html
[root@rs-205 conf.d]# vim /usr/share/nginx/html/index.html
this is test page of realserver02:172.16.60.205
 
[root@rs-205 conf.d]# vim /usr/share/nginx/html/lvs_testpage.html
Test HA Page
   
[root@rs-205 conf.d]# /etc/init.d/nginx start
Starting nginx:                                            [  OK  ]
[root@rs-205 conf.d]# lsof -i:80
COMMAND   PID  USER   FD   TYPE    DEVICE SIZE/OFF NODE NAME
nginx   20839  root    6u  IPv4 289527645      0t0  TCP *:http (LISTEN)
nginx   20840 nginx    6u  IPv4 289527645      0t0  TCP *:http (LISTEN)
   
最后分别访问realserver01和realserver02节点的nginx,:
访问http://172.16.60.204/, 访问结果为"this is test page of realserver01:172.16.60.204"
访问http://172.16.60.204/lvs_testpage.html, 访问结果为"Test HA Page"
 
访问http://172.16.60.205/, 访问结果为"this is test page of realserver02:172.16.60.205"
访问http://172.16.60.205/lvs_testpage.html, 访问结果为"Test HA Page"

5) 配置两台HA节点上转发到自身80端口的页面内容 (两台HA节点操作一致)

由于在ldirectord.cf文件中配置了"fallback=127.0.0.1:80 gate", 即当后端realserver都发生故障时, 客户端的访问请求将转发到LVS的HA节点自身的80端口上
 
[root@ha-master ~]# rpm -ivh http://nginx.org/packages/centos/6/noarch/RPMS/nginx-release-centos-6-0.el6.ngx.noarch.rpm
[root@ha-master ~]# yum install nginx
   
realserver01的nginx配置
[root@ha-master ~]# cd /etc/nginx/conf.d/
[root@ha-master conf.d]# cat default.conf
[root@ha-master conf.d]# >/usr/share/nginx/html/index.html
[root@ha-master conf.d]# vim /usr/share/nginx/html/index.html
Sorry, the access is in maintenance for the time being. Please wait a moment.
 
[root@ha-master conf.d]# /etc/init.d/nginx start
Starting nginx:                                            [  OK  ]
[root@ha-master conf.d]# lsof -i:80
COMMAND   PID  USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
nginx   31944  root    6u  IPv4  91208      0t0  TCP *:http (LISTEN)
nginx   31945 nginx    6u  IPv4  91208      0t0  TCP *:http (LISTEN)
 
访问http://172.16.60.206/  或者 http://172.16.60.207
访问结果为"Sorry, the access is in maintenance for the time being. Please wait a moment."

6) 启动heartbeat服务 (两个HA节点都要操作)

启动heartbeat服务的时候, 就会自带启动ipvsadm 和 ldirectord, 因为在/etc/ha.d/haresources文件里配置了!
需要知道的是: 只有当前提供lvs转发服务(即拥有VIP资源)的一方 才能在启动heartbeat的时候, 自带启动ipvsadm 和 ldirectord!
 
1) 先启动HA主节点的heartbeat
[root@ha-master ~]# /etc/init.d/heartbeat start
Starting High-Availability services: INFO:  Resource is stopped
Done.
 
[root@ha-master ~]# ps -ef|grep heartbeat
root     20886     1  0 15:41 ?        00:00:00 heartbeat: master control process
root     20891 20886  0 15:41 ?        00:00:00 heartbeat: FIFO reader       
root     20892 20886  0 15:41 ?        00:00:00 heartbeat: write: bcast eth0 
root     20893 20886  0 15:41 ?        00:00:00 heartbeat: read: bcast eth0  
root     20894 20886  0 15:41 ?        00:00:00 heartbeat: write: ucast eth0 
root     20895 20886  0 15:41 ?        00:00:00 heartbeat: read: ucast eth0  
root     20896 20886  0 15:41 ?        00:00:00 heartbeat: write: ping_group group1
root     20897 20886  0 15:41 ?        00:00:00 heartbeat: read: ping_group group1
root     20917 20886  0 15:41 ?        00:00:00 /usr/lib64/heartbeat/ipfail
root     20938 17616  0 15:41 pts/0    00:00:00 grep heartbeat
 
heartbeat服务端口默认是694.
[root@ha-master ~]# lsof -i:694
COMMAND     PID USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
heartbeat 20892 root    7u  IPv4  42238      0t0  UDP *:ha-cluster
heartbeat 20893 root    7u  IPv4  42238      0t0  UDP *:ha-cluster
heartbeat 20894 root    7u  IPv4  42244      0t0  UDP *:ha-cluster
heartbeat 20895 root    7u  IPv4  42244      0t0  UDP *:ha-cluster
 
发现ldirectord服务被自带启动了, 说明master节点是当前提供lvs转发服务的一方
[root@ha-master ~]# ps -ef|grep ldirectord
root     21336     1  0 15:41 ?        00:00:00 /usr/bin/perl -w /usr/sbin/ldirectord start
root     21365 17616  0 15:42 pts/0    00:00:00 grep ldirectord
 
[root@ha-master ~]# /etc/init.d/ldirectord status
ldirectord for /etc/ha.d/ldirectord.cf is running with pid: 21336
 
查看master节点,发现master节点当前占有vip资源  (首次启动heartbeat服务后, 需要稍微等待一段时间, vip资源才会出来. 后续再重启或切换时, vip资源就会迅速出现了)
[root@ha-master ~]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:50:9b brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.206/24 brd 172.16.60.255 scope global eth0
    inet 172.16.60.111/24 brd 172.16.60.255 scope global secondary eth0
    inet6 fe80::250:56ff:feac:509b/64 scope link
       valid_lft forever preferred_lft forever
 
master节点当前提供了lvs转发功能, 可以查看到转发效果
[root@ha-master ~]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.204:80             Route   1      0          0        
  -> 172.16.60.205:80             Route   1      0          0 
 
查看master节点的heartbeat日志
[root@ha-master ~]# tail -f /var/log/ha-log
ip-request-resp(default)[21041]:        2018/12/25_15:41:48 received ip-request-resp IPaddr::172.16.60.111 OK yes
ResourceManager(default)[21064]:        2018/12/25_15:41:48 info: Acquiring resource group: ha-master IPaddr::172.16.60.111 ipvsadm ldirectord
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21092]: 2018/12/25_15:41:48 INFO:  Resource is stopped
ResourceManager(default)[21064]:        2018/12/25_15:41:48 info: Running /etc/ha.d/resource.d/IPaddr 172.16.60.111 start
IPaddr(IPaddr_172.16.60.111)[21188]:    2018/12/25_15:41:48 INFO: Adding inet address 172.16.60.111/24 with broadcast address 172.16.60.255 to device eth0
IPaddr(IPaddr_172.16.60.111)[21188]:    2018/12/25_15:41:48 INFO: Bringing device eth0 up
IPaddr(IPaddr_172.16.60.111)[21188]:    2018/12/25_15:41:48 INFO: /usr/libexec/heartbeat/send_arp -i 200 -r 5 -p /var/run/resource-agents/send_arp-172.16.60.111 eth0 172.16.60.111 auto not_used not_used
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21174]: 2018/12/25_15:41:48 INFO:  Success
ResourceManager(default)[21064]:        2018/12/25_15:41:48 info: Running /etc/init.d/ipvsadm  start
ResourceManager(default)[21064]:        2018/12/25_15:41:48 info: Running /etc/init.d/ldirectord  start
 
2) 接着启动HA备份节点的heartbeat
[root@ha-slave ha.d]# /etc/init.d/heartbeat start
Starting High-Availability services: INFO:  Resource is stopped
Done.
 
[root@ha-slave ha.d]# ps -ef|grep heartbeat
root     21703     1  0 15:41 ?        00:00:00 heartbeat: master control process
root     21708 21703  0 15:41 ?        00:00:00 heartbeat: FIFO reader       
root     21709 21703  0 15:41 ?        00:00:00 heartbeat: write: bcast eth0 
root     21710 21703  0 15:41 ?        00:00:00 heartbeat: read: bcast eth0  
root     21711 21703  0 15:41 ?        00:00:00 heartbeat: write: ucast eth0 
root     21712 21703  0 15:41 ?        00:00:00 heartbeat: read: ucast eth0  
root     21713 21703  0 15:41 ?        00:00:00 heartbeat: write: ping_group group1
root     21714 21703  0 15:41 ?        00:00:00 heartbeat: read: ping_group group1
root     21734 21703  0 15:41 ?        00:00:00 /usr/lib64/heartbeat/ipfail
root     21769 19163  0 15:42 pts/0    00:00:00 grep heartbeat
 
[root@ha-slave ha.d]# lsof -i:694
COMMAND     PID USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
heartbeat 21709 root    7u  IPv4 105186      0t0  UDP *:ha-cluster
heartbeat 21710 root    7u  IPv4 105186      0t0  UDP *:ha-cluster
heartbeat 21711 root    7u  IPv4 105192      0t0  UDP *:ha-cluster
heartbeat 21712 root    7u  IPv4 105192      0t0  UDP *:ha-cluster
 
发现ldirectord服务没有被heartbeat自带启动 (因为当前备份节点没有提供lvs转发功能, 即没有接管vip资源)
[root@ha-slave ha.d]# /etc/init.d/ldirectord status
ldirectord is stopped for /etc/ha.d/ldirectord.cf
 
[root@ha-slave ha.d]# ps -ef|grep ldirectord      
root     21822 19163  0 15:55 pts/0    00:00:00 grep ldirectord
 
发现ipvsadm服务也没有被heartbeat自带启动  (因为当前备份节点没有提供lvs转发功能, 即没有接管vip资源)
[root@ha-slave ha.d]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:05:b5 brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.207/24 brd 172.16.60.255 scope global eth0
    inet6 fe80::250:56ff:feac:5b5/64 scope link
       valid_lft forever preferred_lft forever
[root@ha-slave ha.d]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
 
查看HA备份节点的heartbeat日志
[root@ha-slave ha.d]# tail -f /var/log/ha-log  
Dec 25 15:41:37 ha-slave heartbeat: [21734]: info: Starting "/usr/lib64/heartbeat/ipfail" as uid 0  gid 0 (pid 21734)
Dec 25 15:41:38 ha-slave heartbeat: [21703]: info: Status update for node ha-master: status active
harc(default)[21737]:   2018/12/25_15:41:38 info: Running /etc/ha.d//rc.d/status status
Dec 25 15:41:42 ha-slave ipfail: [21734]: info: Status update: Node ha-master now has status active
Dec 25 15:41:44 ha-slave ipfail: [21734]: info: Asking other side for ping node count.
Dec 25 15:41:47 ha-slave ipfail: [21734]: info: No giveup timer to abort.
Dec 25 15:41:48 ha-slave heartbeat: [21703]: info: remote resource transition completed.
Dec 25 15:41:48 ha-slave heartbeat: [21703]: info: remote resource transition completed.
Dec 25 15:41:48 ha-slave heartbeat: [21703]: info: Initial resource acquisition complete (T_RESOURCES(us))
Dec 25 15:41:48 ha-slave heartbeat: [21754]: info: No local resources [/usr/share/heartbeat/Resourc
 
访问使用vip地址访问, 即:
访问http://172.16.60.111/, 结果为"this is test page of realserver01:172.16.60.204" 或者 "this is test page of realserver02:172.16.60.205"
访问http://172.16.60.111/lvs_testpage.html, 结果为"Test HA Page"
 
温馨提示:
下面是两个常用的ipvsadm 关于查看lvs状态的命令
======================================
查看lvs的连接状态命令
[root@ha-master ~]# ipvsadm  -l  --stats
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port               Conns   InPkts  OutPkts  InBytes OutBytes
  -> RemoteAddress:Port
TCP  172.16.60.111:http                0        0        0        0        0
  -> 172.16.60.204:http                0        0        0        0        0
  -> 172.16.60.205:http                0        0        0        0        0
 
说明：
Conns    (connections scheduled)  已经转发过的连接数
InPkts   (incoming packets)       入包个数
OutPkts  (outgoing packets)       出包个数
InBytes  (incoming bytes)         入流量（字节） 
OutBytes (outgoing bytes)         出流量（字节）
 
======================================
查看lvs的速率
[root@ha-master ~]# ipvsadm   -l  --rate
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port                 CPS    InPPS   OutPPS    InBPS   OutBPS
  -> RemoteAddress:Port
TCP  172.16.60.111:http                0        0        0        0        0
  -> 172.16.60.204:http                0        0        0        0        0
  -> 172.16.60.205:http                0        0        0        0        0
 
说明：
CPS      (current connection rate)   每秒连接数
InPPS    (current in packet rate)    每秒的入包个数
OutPPS   (current out packet rate)   每秒的出包个数
InBPS    (current in byte rate)      每秒入流量（字节）
OutBPS   (current out byte rate)     每秒入流量（字节）
 
======================================
上面的两台HA节点均只有一个网卡设备eth0,  如果有两块网卡, 比如还有一个eth1, 则可以将这个eth1作为heartbeat交叉线直连的设备,
即HA主备两台机器之间使用一根串口直连线缆eth1进行连接.
比如:
HA主节点   172.16.60.206(eth0), 10.0.11.21(eth1, heartbeat交叉线直连)
HA备节点   172.16.60.207(eth0), 10.0.11.22(eth1, heartbeat交叉线直连)
 
这样比起只有一个eth0, 只需要在ha.cf文件中多加下面一行 (其他的操作配置都不用变!)
ping_group group1 10.0.11.21 10.0.11.22       //多加这一行
ping_group group1 172.16.60.204 172.16.60.205

7) 故障转移切换测试

1) 先关闭HA主节点的heartbeat
[root@ha-master ~]# /etc/init.d/heartbeat stop
Stopping High-Availability services: Done.
 
[root@ha-master ~]# ps -ef|grep heartbeat
root     21625 17616  0 16:03 pts/0    00:00:00 grep heartbeat
 
发现关闭heartbeat服务后, 主节点的ipvsadm 和 ldirectord都会被自带关闭, VIP资源也被转移走了, 即当前master节点不提供lvs转发服务
[root@ha-master ~]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:50:9b brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.206/24 brd 172.16.60.255 scope global eth0
    inet6 fe80::250:56ff:feac:509b/64 scope link
       valid_lft forever preferred_lft forever
 
[root@ha-master ~]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
 
[root@ha-master ~]# ps -ef|grep ldirectord
root     21630 17616  0 16:03 pts/0    00:00:00 grep ldirectord
 
查看此时HA主节点的heartbeat日志
[root@ha-master ~]# tail -1000 /var/log/ha-log
........
Dec 25 16:02:38 ha-master heartbeat: [20886]: info: Heartbeat shutdown in progress. (20886)
Dec 25 16:02:38 ha-master heartbeat: [21454]: info: Giving up all HA resources.
ResourceManager(default)[21467]:        2018/12/25_16:02:38 info: Releasing resource group: ha-master IPaddr::172.16.60.111 ipvsadm ldirectord
ResourceManager(default)[21467]:        2018/12/25_16:02:38 info: Running /etc/init.d/ldirectord  stop
ResourceManager(default)[21467]:        2018/12/25_16:02:38 info: Running /etc/init.d/ipvsadm  stop
ResourceManager(default)[21467]:        2018/12/25_16:02:38 info: Running /etc/ha.d/resource.d/IPaddr 172.16.60.111 stop
IPaddr(IPaddr_172.16.60.111)[21563]:    2018/12/25_16:02:38 INFO: IP status = ok, IP_CIP=
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21549]: 2018/12/25_16:02:38 INFO:  Success
 
接着查看HA备份节点的情况, 发现VIP已将已经切换到备份节点这边了, 说明当前备份节点提供lvs转发服务, 则备份节点的ipvsadm 和 ldirectord也被自带启动了
[root@ha-slave ha.d]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:05:b5 brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.207/24 brd 172.16.60.255 scope global eth0
    inet 172.16.60.111/24 brd 172.16.60.255 scope global secondary eth0
    inet6 fe80::250:56ff:feac:5b5/64 scope link
       valid_lft forever preferred_lft forever
 
[root@ha-slave ha.d]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.204:80             Route   1      0          0        
  -> 172.16.60.205:80             Route   1      0          0 
 
[root@ha-slave ha.d]# ps -ef|grep ldirectord
root     22203     1  0 16:02 ?        00:00:01 /usr/bin/perl -w /usr/sbin/ldirectord start
root     22261 19163  0 16:07 pts/0    00:00:00 grep ldirectord
 
查看此时HA备份节点的heartbeat日志
[root@ha-slave ha.d]# tail -1000 /var/log/ha-log
...........
harc(default)[21887]:   2018/12/25_16:02:39 info: Running /etc/ha.d//rc.d/status status
mach_down(default)[21904]:      2018/12/25_16:02:39 info: Taking over resource group IPaddr::172.16.60.111
ResourceManager(default)[21931]:        2018/12/25_16:02:39 info: Acquiring resource group: ha-master IPaddr::172.16.60.111 ipvsadm ldirectord
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21959]: 2018/12/25_16:02:39 INFO:  Resource is stopped
ResourceManager(default)[21931]:        2018/12/25_16:02:39 info: Running /etc/ha.d/resource.d/IPaddr 172.16.60.111 start
IPaddr(IPaddr_172.16.60.111)[22055]:    2018/12/25_16:02:39 INFO: Adding inet address 172.16.60.111/24 with broadcast address 172.16.60.255 to device eth0
IPaddr(IPaddr_172.16.60.111)[22055]:    2018/12/25_16:02:39 INFO: Bringing device eth0 up
IPaddr(IPaddr_172.16.60.111)[22055]:    2018/12/25_16:02:39 INFO: /usr/libexec/heartbeat/send_arp -i 200 -r 5 -p /var/run/resource-agents/send_arp-172.16.60.111 eth0 172.16.60.111 auto not_used not_used
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[22041]: 2018/12/25_16:02:39 INFO:  Success
ResourceManager(default)[21931]:        2018/12/25_16:02:39 info: Running /etc/init.d/ipvsadm  start
ResourceManager(default)[21931]:        2018/12/25_16:02:39 info: Running /etc/init.d/ldirectord  start
mach_down(default)[21904]:      2018/12/25_16:02:39 info: /usr/share/heartbeat/mach_down: nice_failback: foreign resources acquired
mach_down(default)[21904]:      2018/12/25_16:02:39 info: mach_down takeover complete for node ha-master.
 
2) 然后在重新启动HA主节点的heartbeat服务
由于在ha.cf文件中配置了"auto_failback on "参数, 所以当主节点恢复后, 会将VIP资源自动抢占回来并替换备份节点重新接管lvs转发服务.
主节点的heartbeat恢复后, ipvsadm 和 ldirectord也会被重新启动
 
[root@ha-master ~]# /etc/init.d/heartbeat start
Starting High-Availability services: INFO:  Resource is stopped
Done.
 
[root@ha-master ~]# ps -ef|grep heartbeat
root     21778     1  0 16:12 ?        00:00:00 heartbeat: master control process
root     21783 21778  0 16:12 ?        00:00:00 heartbeat: FIFO reader       
root     21784 21778  0 16:12 ?        00:00:00 heartbeat: write: bcast eth0 
root     21785 21778  0 16:12 ?        00:00:00 heartbeat: read: bcast eth0  
root     21786 21778  0 16:12 ?        00:00:00 heartbeat: write: ucast eth0 
root     21787 21778  0 16:12 ?        00:00:00 heartbeat: read: ucast eth0  
root     21788 21778  0 16:12 ?        00:00:00 heartbeat: write: ping_group group1
root     21789 21778  0 16:12 ?        00:00:00 heartbeat: read: ping_group group1
root     21809 21778  0 16:12 ?        00:00:00 /usr/lib64/heartbeat/ipfail
root     21812 21778  0 16:12 ?        00:00:00 heartbeat: master control process
root     21825 21812  0 16:12 ?        00:00:00 /bin/sh /usr/share/heartbeat/ResourceManager takegroup IPaddr::172.16.60.111 ipvsadm ldirectord
root     21949 21935  0 16:12 ?        00:00:00 /bin/sh /usr/lib/ocf/resource.d//heartbeat/IPaddr start
root     21956 17616  0 16:12 pts/0    00:00:00 grep heartbeat
 
[root@ha-master ~]# lsof -i:694
COMMAND     PID USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
heartbeat 21784 root    7u  IPv4  46306      0t0  UDP *:ha-cluster
heartbeat 21785 root    7u  IPv4  46306      0t0  UDP *:ha-cluster
heartbeat 21786 root    7u  IPv4  46312      0t0  UDP *:ha-cluster
heartbeat 21787 root    7u  IPv4  46312      0t0  UDP *:ha-cluster
 
[root@ha-master ~]# ps -ef|grep ldirectord    
root     22099     1  1 16:12 ?        00:00:00 /usr/bin/perl -w /usr/sbin/ldirectord start
root     22130 17616  0 16:12 pts/0    00:00:00 grep ldirectord
 
[root@ha-master ~]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:50:9b brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.206/24 brd 172.16.60.255 scope global eth0
    inet 172.16.60.111/24 brd 172.16.60.255 scope global secondary eth0
    inet6 fe80::250:56ff:feac:509b/64 scope link
       valid_lft forever preferred_lft forever
 
[root@ha-master ~]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.204:80             Route   1      0          0        
  -> 172.16.60.205:80             Route   1      1          0
 
查看此时HA主节点的heartbeat日志
[root@ha-master ~]# tail -1000 /var/log/ha-log
........
ResourceManager(default)[21825]:        2018/12/25_16:12:12 info: Acquiring resource group: ha-master IPaddr::172.16.60.111 ipvsadm ldirectord
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21853]: 2018/12/25_16:12:13 INFO:  Resource is stopped
ResourceManager(default)[21825]:        2018/12/25_16:12:13 info: Running /etc/ha.d/resource.d/IPaddr 172.16.60.111 start
IPaddr(IPaddr_172.16.60.111)[21949]:    2018/12/25_16:12:13 INFO: Adding inet address 172.16.60.111/24 with broadcast address 172.16.60.255 to device eth0
IPaddr(IPaddr_172.16.60.111)[21949]:    2018/12/25_16:12:13 INFO: Bringing device eth0 up
IPaddr(IPaddr_172.16.60.111)[21949]:    2018/12/25_16:12:13 INFO: /usr/libexec/heartbeat/send_arp -i 200 -r 5 -p /var/run/resource-agents/send_arp-172.16.60.111 eth0 172.16.60.111 auto not_used not_used
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[21935]: 2018/12/25_16:12:13 INFO:  Success
ResourceManager(default)[21825]:        2018/12/25_16:12:13 info: Running /etc/init.d/ipvsadm  start
ResourceManager(default)[21825]:        2018/12/25_16:12:13 info: Running /etc/init.d/ldirectord  start
 
再观察此时HA备份节点的情况, 发现VIP资源在主节点的heartbeat恢复后就被主节点抢占回去了, 即此时备份节点没有vip资源, 也就不提供lvs转发服务了,
则备份节点的ipvsadm 和 ldirectord服务也会被关闭
[root@ha-slave ha.d]# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:ac:05:b5 brd ff:ff:ff:ff:ff:ff
    inet 172.16.60.207/24 brd 172.16.60.255 scope global eth0
    inet6 fe80::250:56ff:feac:5b5/64 scope link
       valid_lft forever preferred_lft forever
 
[root@ha-slave ha.d]# ps -ef|grep ldirectord    
root     22516 19163  0 16:14 pts/0    00:00:00 grep ldirectord
 
[root@ha-slave ha.d]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
 
查看此时HA备份节点的heartbeat日志
[root@ha-slave ha.d]# tail -1000 /var/log/ha-log
.......
ResourceManager(default)[22342]:        2018/12/25_16:12:12 info: Releasing resource group: ha-master IPaddr::172.16.60.111 ipvsadm ldirectord
ResourceManager(default)[22342]:        2018/12/25_16:12:12 info: Running /etc/init.d/ldirectord  stop
ResourceManager(default)[22342]:        2018/12/25_16:12:12 info: Running /etc/init.d/ipvsadm  stop
ResourceManager(default)[22342]:        2018/12/25_16:12:12 info: Running /etc/ha.d/resource.d/IPaddr 172.16.60.111 stop
IPaddr(IPaddr_172.16.60.111)[22438]:    2018/12/25_16:12:12 INFO: IP status = ok, IP_CIP=
/usr/lib/ocf/resource.d//heartbeat/IPaddr(IPaddr_172.16.60.111)[22424]: 2018/12/25_16:12:12 INFO:  Success
Dec 25 16:12:12 ha-slave heartbeat: [22329]: info: foreign HA resource release completed (standby).
 
在上面HA主备节点故障切换的过程中, 客户端访问http://172.16.60.111/都是不受影响的, 即对客户端访问来说是无感知的故障切换, 实现了lvs代理层的高可用!
 
3) 先后关闭两台realserver节点中的nginx, 然后观察lvs的转发情况
[root@ha-master ~]# ipvsadm -Ln               
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.204:80             Route   1      0          0        
  -> 172.16.60.205:80             Route   1      0          2  
 
先关闭rs-204的nginx服务
[root@rs-204 ~]# /etc/init.d/nginx stop
Stopping nginx:                                            [  OK  ]
[root@rs-204 ~]# lsof -i:80
[root@rs-204 ~]#
 
rs-205的nginx保留
[root@rs-205 ~]# ps -ef|grep nginx
root      5211     1  0 15:45 ?        00:00:00 nginx: master process /usr/sbin/nginx -c /etc/nginx/nginx.conf
nginx     5212  5211  0 15:45 ?        00:00:00 nginx: worker process                  
root      5313  4852  0 16:19 pts/0    00:00:00 grep nginx
 
查看lvs转发情况
[root@ha-master ~]# ipvsadm -Ln               
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.205:80             Route   1      0          2        
 
这时候访问http://172.16.60.111, 结果是"this is test page of realserver02:172.16.60.205"
 
接着启动rs-204的nginx, 关闭rs-205的nginx
[root@rs-204 ~]# /etc/init.d/nginx start
Starting nginx:                                            [  OK  ]
[root@rs-204 ~]# lsof -i:80            
COMMAND  PID  USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
nginx   4883  root    6u  IPv4 143621      0t0  TCP *:http (LISTEN)
nginx   4884 nginx    6u  IPv4 143621      0t0  TCP *:http (LISTEN)
 
关闭rs-205的nginx
[root@rs-205 ~]# /etc/init.d/nginx stop
Stopping nginx:                                            [  OK  ]
[root@rs-205 ~]# lsof -i:80
[root@rs-205 ~]#
 
查看lvs转发情况
[root@ha-master ~]# ipvsadm -Ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 172.16.60.204:80             Route   1      0          0 
 
这时候访问http://172.16.60.111, 结果是"this is test page of realserver01:172.16.60.204"
 
然后把rs-204 和 rs-205两个节点的nginx都关闭
[root@rs-204 ~]# /etc/init.d/nginx stop
Stopping nginx:                                            [  OK  ]
[root@rs-205 ~]# /etc/init.d/nginx stop
Stopping nginx:                                            [  OK  ]
 
查看lvs转发情况
[root@ha-master ~]# ipvsadm -Ln               
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  172.16.60.111:80 wlc persistent 600
  -> 127.0.0.1:80                 Local   1      0          0  
 
这时候访问http://172.16.60.111, 结果是"Sorry, the access is in maintenance for the time being. Please wait a moment."
 
上面可知, 在realserver节点发生故障后, 会从lvs集群中踢出来, 待realserver节点恢复后会再次重新加入到lvs集群中
这是因为在ldirectord.cf文件中配置了"quiescent=no "参数 , 这样就实现了代理节点的高可用! 

posted @ 2020-01-03 10:13 斑呢个斑马阅读(252) 评论(0) 编辑收藏举报

刷新页面返回顶部

斑呢个斑马

LVS+Heartbeat 高可用集群方案操作记录

公告