Percona-XtraDB-Cluster+haproxy+keepalived编译安装

系统：centos7

专注于为中小企业提供网站建设、网站制作服务,电脑端+手机端+微信端的三站合一,更高效的管理,为中小企业当涂免费做网站提供优质的服务。我们立足成都，凝聚了一批互联网行业人才，有力地推动了1000多家企业的稳健成长，帮助中小企业通过网站建设实现规模扩充和转变。

192.168.1.55 MySQL 端口3306 Percona-XtraDB-Cluster 主节点

192.168.1.56 mysql端口3307 Percona-XtraDB-Cluster+haproxy+keepalived

192.168.1.57 MySQL端口3307 Percona-XtraDB-Cluster+haproxy+keepalived

192.168.4.58/21 vip ip

软件版本：Percona-XtraDB-Cluster-5.7.16-27.19 percona-xtrabackup-2.4.5 

                  haproxy-1.7.0   keepalived-1.3.2

一、linux 系统限制配置

1、关闭系统防火墙  

systemctl stop firewalld.service 关闭防火墙
systemctl disable firewalld.service  禁用防火墙

2、关闭SElinux
sed -i 's/SELINUX=.*/SELINUX=disabled/g' /etc/selinux/config
setenforce 0 selinux 立即生效

二、系统安装约定

源码包编译安装位置：/usr/local/软件名字

数据目录 /apps/data

三、下在源码安装包

下载 Percona-XtraDB-Cluster-5.7.16-27.19

wget -P /usr/local/src https://www.percona.com/downloads/Percona-XtraDB-Cluster-57/Percona-XtraDB-Cluster-5.7.16-27.19/rel-5.7.16/source/tarball/Percona-XtraDB-Cluster-5.7.16-27.19.tar.gz

下载percona-xtrabackup-2.4.5

wget -P /usr/local/src https://www.percona.com/downloads/XtraBackup/Percona-XtraBackup-2.4.5/source/tarball/percona-xtrabackup-2.4.5.tar.gz

下载boost_1_59_0

mkdir -p /usr/local/boost
wget -P /usr/local/boost https://nchc.dl.sourceforge.net/project/boost/boost/1.59.0/boost_1_59_0.tar.gz

下载haproxy-1.7.0

 wget -P /usr/local/src http://www.haproxy.org/download/1.7/src/haproxy-1.7.0.tar.gz

下载keepalived-1.3.2

wget -P /usr/local/src http://www.keepalived.org/software/keepalived-1.3.2.tar.gz

下载automake-1.15

wget -P /usr/local/src ftp://ftp.gnu.org/gnu/automake/automake-1.15.tar.gz

四、安装编译环境及编译依赖

1、yum -y install epel-release
2、卸载旧版mysql或mariadb
yum -y remove mariadb* mysql*
3、安装编译环境
yum -y install  libtool ncurses-devel libgcrypt-devel  libev-devel \
git scons  gcc  gcc-c++  openssl  check  cmake  bison boost-devel  \
asio-devel libaio-devel  ncurses-devel  readline-devel  pam-devel socat \
libaio  automake autoconf vim redhat-lsb check-devel curl  curl-devel xinetd
4、安装keepalived 依赖 （192.168.1.56 192.168.1.57）
yum -y install libnl-devel openssl-devel libnfnetlink-devel ipvsadm \
popt-devel libnfnetlink kernel-devel popt-static iptraf

五、编译安装 Percona-XtraDB-Cluster-5.7.16-27.19 percona-xtrabackup-2.4.5

1、解压percona-xtrabackup-2.4.5.tar.gz

tar -xvf /usr/local/src/percona-xtrabackup-2.4.5.tar.gz -C /usr/local/src/

2、解压Percona-XtraDB-Cluster-5.7.16-27.19.tar.gz

tar -xvf /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19.tar.gz -C /usr/local/src/

3、编译percona-xtrabackup-2.4.5

cd /usr/local/src/percona-xtrabackup-2.4.5/
cmake ./ -DBUILD_CONFIG=xtrabackup_release \
         -DWITH_MAN_PAGES=OFF \
         -DDOWNLOAD_BOOST=1 \
         -DWITH_BOOST="/usr/local/boost" 
make -j4 && make install

ln -sf /usr/local/xtrabackup/bin/* /usr/sbin/

4、编译安装Percona-XtraDB-Cluster-5.7.16-27.19

4.1、创建mysql用户名及文件夹

mkdir /apps/{run,log}/mysqld -p
useradd mysql -s /sbin/nologin -M
chown -R mysql:mysql /apps

4.2、编译garbd,libgalera_smm.so

cd /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19
cd "percona-xtradb-cluster-galera"
revno值 cat GALERA-REVISION
scons -j4 psi=1 --config=force  revno="b98f92f"  boost_pool=0 libgalera_smm.so
scons -j4 --config=force revno="b98f92f" garb/garbd
创建pxc安装目录
mkdir -p /usr/local/Percona-XtraDB-Cluster/{bin,lib}
拷贝编译结果文件到pxc安装目录
cp garb/garbd /usr/local/Percona-XtraDB-Cluster/bin
cp libgalera_smm.so /usr/local/Percona-XtraDB-Cluster/lib

4.3、编译 Percona-XtraDB-Cluster-5.7.16-27.19

-DMYSQL_SERVER_SUFFIX 值得获取
cd /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19
WSREP_VERSION="$(grep WSREP_INTERFACE_VERSION wsrep/wsrep_api.h | cut -d '"' -f2).$(grep 'SET(WSREP_PATCH_VERSION'  "cmake/wsrep.cmake" | cut -d '"' -f2)"
echo $WSREP_VERSION
-DCOMPILATION_COMMENT 值获取
cd /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19
source VERSION
MYSQL_VERSION="$MYSQL_VERSION_MAJOR.$MYSQL_VERSION_MINOR.$MYSQL_VERSION_PATCH"
 echo $MYSQL_VERSION
  REVISION="$(cd "$SOURCEDIR"; grep '^short: ' Docs/INFO_SRC |sed -e 's/short: //')"
 -DCOMPILATION_COMMENT 值
 Percona XtraDB Cluster binary (GPL) $MYSQL_VERSION-$WSREP_VERSION Revision $REVISION
 编译Percona-XtraDB-Cluster-5.7.16-27.19 
cd /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19

cmake ./ -DBUILD_CONFIG=mysql_release \
            -DCMAKE_BUILD_TYPE=RelWithDebInfo\
            -DWITH_EMBEDDED_SERVER=OFF \
            -DFEATURE_SET=community \
            -DENABLE_DTRACE=OFF \
            -DWITH_SSL=system -DWITH_ZLIB=system \
            -DCMAKE_INSTALL_PREFIX="/usr/local/Percona-XtraDB-Cluster" \
            -DMYSQL_DATADIR="/apps/data" \
            -DMYSQL_SERVER_SUFFIX="27.19" \
            -DWITH_INNODB_DISALLOW_WRITES=ON \
            -DWITH_WSREP=ON \
            -DWITH_UNIT_TESTS=0 \
            -DWITH_READLINE=system \
            -DWITHOUT_TOKUDB=ON \
	    -DCOMPILATION_COMMENT="Percona XtraDB Cluster binary (GPL) 5.7.16-27.19, Revision bec0879" \
            -DWITH_PAM=ON \
            -DWITH_INNODB_MEMCACHED=ON \
            -DDOWNLOAD_BOOST=1 \
            -DWITH_BOOST="/usr/local/boost" \
            -DWITH_SCALABILITY_METRICS=ON 
  make -j4 && make install
  拷贝文件到安装目录
  cp -R /usr/local/src/Percona-XtraDB-Cluster-5.7.16-27.19/percona-xtradb-cluster-tests /usr/local/Percona-XtraDB-Cluster/
  创建软连接 mysql 系统环境
  ln -sf /usr/local/Percona-XtraDB-Cluster/bin/* /usr/sbin/
  拷贝启动文件到系统启动目录
  cp -R /usr/local/Percona-XtraDB-Cluster/support-files/mysql.server /etc/init.d/mysqld
  拷贝监控文件到xinetd管理目录
  cp -R /usr/local/Percona-XtraDB-Cluster/xinetd.d/mysqlchk /etc/xinetd.d/
  加入动态库
  echo "/usr/local/Percona-XtraDB-Cluster/lib" >> /etc/ld.so.conf
  ldconfig

4.4 配置MySQL健康检查

vim /etc/xinetd.d/mysqlchk
server          = /usr/bin/clustercheck
修改为如果不修过可以cp /usr/local/Percona-XtraDB-Cluster/bin/clustercheck
到 /usr/bin/
server          = /usr/sbin/clustercheck
vim /etc/services
添加
mysqlchk 9200/tcp # mysqlchk

4.5my.cnf 配置

192.168.1.55
vim /etc/my.cnf
# Template my.cnf for PXC
# Edit to your requirements.
[client]
socket=/apps/data/mysql.sock
port=3306
[mysqld]
user=mysql
server-id=10055
datadir=/apps/data
socket=/apps/data/mysql.sock
log-error=/apps/log/mysqld/mysqld.log
pid-file=/apps/run/mysqld/mysqld.pid
log-bin
log_slave_updates
expire_logs_days=7
port=3306
character_set_server = utf8
# Disabling symbolic-links is recommended to prevent assorted security risks
symbolic-links=0

# Path to Galera library
wsrep_provider=/usr/local/Percona-XtraDB-Cluster/lib/libgalera_smm.so

# Cluster connection URL contains IPs of nodes
#If no IP is found, this implies that a new cluster needs to be created,
#in order to do that you need to bootstrap this node
wsrep_cluster_address='gcomm://192.168.1.55,192.168.1.56,192.168.1.57'
#wsrep_cluster_address=gcomm://

# In order for Galera to work correctly binlog format should be ROW
binlog_format=ROW

# MyISAM storage engine has only experimental support
default_storage_engine=InnoDB


wsrep_sst_receive_address=192.168.1.55:6010

wsrep_node_incoming_address=192.168.1.55

wsrep_provider_options = "gmcast.listen_addr=tcp://192.168.1.55;ist.recv_addr=192.168.1.55:6020;"

# Slave thread to use
wsrep_slave_threads= 8

wsrep_log_conflicts

# This changes how InnoDB autoincrement locks are managed and is a requirement for Galera
innodb_autoinc_lock_mode=2

# Node IP address
wsrep_node_address=192.168.1.55

# Cluster name
wsrep_cluster_name=pxc-cluster

#If wsrep_node_name is not specified,  then system hostname will be used
wsrep_node_name=pxc-cluster-node-1

#pxc_strict_mode allowed values: DISABLED,PERMISSIVE,ENFORCING,MASTER
pxc_strict_mode=ENFORCING

# SST method
wsrep_sst_method=xtrabackup-v2
#wsrep_sst_method=rsync

#Authentication for SST method # 修改自己的创建的账号、密码
wsrep_sst_auth="sstuser:456789"

192.168.1.56
vim /etc/my.cnf
# Template my.cnf for PXC
# Edit to your requirements.
[client]
socket=/apps/data/mysql.sock
port=3307
[mysqld]
user=mysql
server-id=10056
datadir=/apps/data
socket=/apps/data/mysql.sock
log-error=/apps/log/mysqld/mysqld.log
pid-file=/apps/run/mysqld/mysqld.pid
log-bin
log_slave_updates
expire_logs_days=7
port=3307
character_set_server = utf8
# Disabling symbolic-links is recommended to prevent assorted security risks
symbolic-links=0

# Path to Galera library
wsrep_provider=/usr/local/Percona-XtraDB-Cluster/lib/libgalera_smm.so

# Cluster connection URL contains IPs of nodes
#If no IP is found, this implies that a new cluster needs to be created,
#in order to do that you need to bootstrap this node
wsrep_cluster_address='gcomm://192.168.1.55,192.168.1.56,192.168.1.57'
#wsrep_cluster_address=gcomm://

# In order for Galera to work correctly binlog format should be ROW
binlog_format=ROW

# MyISAM storage engine has only experimental support
default_storage_engine=InnoDB


wsrep_sst_receive_address=192.168.1.56:6010

wsrep_node_incoming_address=192.168.1.56

wsrep_provider_options = "gmcast.listen_addr=tcp://192.168.1.56;ist.recv_addr=192.168.1.56:6020;"

# Slave thread to use
wsrep_slave_threads= 8

wsrep_log_conflicts

# This changes how InnoDB autoincrement locks are managed and is a requirement for Galera
innodb_autoinc_lock_mode=2

# Node IP address
wsrep_node_address=192.168.1.56


# Cluster name
wsrep_cluster_name=pxc-cluster

#If wsrep_node_name is not specified,  then system hostname will be used
wsrep_node_name=pxc-cluster-node-2

#pxc_strict_mode allowed values: DISABLED,PERMISSIVE,ENFORCING,MASTER
pxc_strict_mode=ENFORCING

# SST method
wsrep_sst_method=xtrabackup-v2
#wsrep_sst_method=rsync

#Authentication for SST method # 修改自己的创建的账号、密码
wsrep_sst_auth="sstuser:456789"

192.168.1.57
vim /etc/my.cnf
# Template my.cnf for PXC
# Edit to your requirements.
[client]
socket=/apps/data/mysql.sock
port=3307
[mysqld]
user=mysql
server-id=10057
datadir=/apps/data
socket=/apps/data/mysql.sock
log-error=/apps/log/mysqld/mysqld.log
pid-file=/apps/run/mysqld/mysqld.pid
log-bin
log_slave_updates
expire_logs_days=7
port=3307
character_set_server = utf8
# Disabling symbolic-links is recommended to prevent assorted security risks
symbolic-links=0

# Path to Galera library
wsrep_provider=/usr/local/Percona-XtraDB-Cluster/lib/libgalera_smm.so

# Cluster connection URL contains IPs of nodes
#If no IP is found, this implies that a new cluster needs to be created,
#in order to do that you need to bootstrap this node
wsrep_cluster_address='gcomm://192.168.1.55,192.168.1.56,192.168.1.57'
#wsrep_cluster_address=gcomm://

# In order for Galera to work correctly binlog format should be ROW
binlog_format=ROW

# MyISAM storage engine has only experimental support
default_storage_engine=InnoDB


wsrep_sst_receive_address=192.168.1.57:6010

wsrep_node_incoming_address=192.168.1.57

wsrep_provider_options = "gmcast.listen_addr=tcp://192.168.1.57;ist.recv_addr=192.168.1.57:6020;"

# Slave thread to use
wsrep_slave_threads= 8

wsrep_log_conflicts

# This changes how InnoDB autoincrement locks are managed and is a requirement for Galera
innodb_autoinc_lock_mode=2

# Node IP address
wsrep_node_address=192.168.1.57

# Cluster name
wsrep_cluster_name=pxc-cluster

#If wsrep_node_name is not specified,  then system hostname will be used
wsrep_node_name=pxc-cluster-node-3

#pxc_strict_mode allowed values: DISABLED,PERMISSIVE,ENFORCING,MASTER
pxc_strict_mode=ENFORCING

# SST method
wsrep_sst_method=xtrabackup-v2
#wsrep_sst_method=rsync

#Authentication for SST method
wsrep_sst_auth="sstuser:456789" # 修改自己的创建的账号、密码

4.6 初始化数据库

192.168.1.55
初始化数据
mysqld --initialize --user=mysql  --datadir="/apps/data"
启动数据库
/etc/init.d/mysqld start --wsrep-cluster-address="gcomm://"

关闭数据库 MySQL5.7以后都要密码才能进去
/etc/init.d/mysqld stop
安全模式期待MySQL修改密码
mysqld_safe --wsrep-cluster-address=gcomm://  --user=mysql --skip-grant-tables --skip-networking &
use mysql；
update user set authentication_string=Password('123456') where user='root';
flush privileges;
shutdown
exit 退出MySQL
ps -ef | grep mysql
直到MySQL完全关闭 再次启动MySQL
/etc/init.d/mysqld start --wsrep-cluster-address="gcomm://"
进入MySQL 创建同步账号及修改root密码
mysql -u root -p123456
创建MySQL同步账号及密码

GRANT ALL PRIVILEGES  ON *.* TO sstuser@'%' IDENTIFIED BY '456789'; FLUSH PRIVILEGES;
GRANT ALL PRIVILEGES  ON *.* TO sstuser@'localhost' IDENTIFIED BY '456789'; FLUSH PRIVILEGES;
修改root 远程能够连接
GRANT ALL PRIVILEGES ON *.* TO root@'%' IDENTIFIED BY '123456'; FLUSH PRIVILEGES;
创建健康检查数据库账号密码

GRANT PROCESS ON *.* TO 'clustercheckuser'@'localhost' IDENTIFIED BY 'clustercheckpassword!';
如果使用自己账号密码请修改
vim /usr/sbin/clustercheck 对应的账号密码
192.168.1.56 192.168.1.57 其它节点直接启动
service mysqld start 

 重启第一个节点 192.168.1.55
 service mysqld restart 出现错误
 /etc/init.d/mysqld bootstrap-pxc 进行启动
 
 启动完成登陆MySQL
 mysql -u root -p123456
 查看MySQL 节点数
 mysql> show global status like 'wsrep_cluster_size';
+--------------------+-------+
| Variable_name      | Value |
+--------------------+-------+
| wsrep_cluster_size | 3     |
+--------------------+-------+
1 row in set (0.00 sec)

mysql> show global status like 'wsrep_incoming_addresses';
+--------------------------+-------------------------------------------------------+
| Variable_name            | Value                                                 |
+--------------------------+-------------------------------------------------------+
| wsrep_incoming_addresses | 192.168.1.57:3307,192.168.1.56:3307,192.168.1.55:3306 |
+--------------------------+-------------------------------------------------------+
1 row in set (0.00 sec)
创建测试数据看看是否同步
mysql> create database test;
Query OK, 1 row affected (0.01 sec)

4.7 数据库健康检查启动

systemctl start  xinetd.service
查看9200端口是否打开
netstat -anlp | grep 9200
tcp6       0      0 :::9200                 :::*                    LISTEN      73323/xinetd 
测试端口是否正常
telnet 192.168.1.55 9200

Escape character is '^]'.
HTTP/1.1 200 OK
Content-Type: text/plain
Connection: close
Content-Length: 40

Percona XtraDB Cluster Node is synced.
Connection closed by foreign host.
返回503 首先执行 
/usr/sbin/clustercheck
是否返回200 如果不是检查链接数据库的账号密码
返回200正常 就请修改 vim /etc/xinetd.d/mysqlchk
user root
重启systemctl restart  xinetd.service
再次测试如果返回200正常就ok 还是报错就请查看系统日志修复

六、编译安装 haproxy (192.168.1.56,192.168.1.57)

安装 haproxy 日志记录支持
yum -y install rsyslog
解压 haproxy-1.7.0.tar.gz 
tar -xvf /usr/local/src/haproxy-1.7.0.tar.gz -C /usr/local/src/
cd /usr/local/src/haproxy-1.7.0
make TARGET=linux31 PREFIX=/usr/local/haproxy
make install PREFIX=/usr/local/haproxy
创建文件夹haproxy
mkdir -pv /usr/local/haproxy/{conf,run,log}

创建haproxy 启动脚本
vim /etc/init.d/haproxy

#! /bin/bash
# chkconfig: - 85 15
# description: haproxy is a World Wide Web server. It is used to serve
PROGDIR=/usr/local/haproxy
PROGNAME=haproxy
DAEMON=$PROGDIR/sbin/$PROGNAME
CONFIG=$PROGDIR/conf/$PROGNAME.conf
PIDFILE=$PROGDIR/run/$PROGNAME.pid
DESC="HAProxy daemon"
SCRIPTNAME=/usr/local/haproxy/init.d/$PROGNAME
# Gracefully exit if the package has been removed.
test -x $DAEMON || exit 0
start()
 {
    echo -n "Starting $DESC: $PROGNAME"
   $DAEMON -f $CONFIG
   echo "."
        }
stop()
 {  echo -n "Stopping $DESC: $PROGNAME"
    cat $PIDFILE | xargs kill
    echo "."
}

reload()
 { echo -n "reloading $DESC: $PROGNAME"
   $DAEMON -f $CONFIG -p $PIDFILE -sf $(cat $PIDFILE)
}
case "$1" in
  start)
  start
  ;;
 stop)
  stop
   ;;
  restart)
   stop
   start
    ;; 
 reload)
 reload
   ;;
*)
 echo "Usage: $SCRIPTNAME {start|stop|restart|reload}" >&2
 exit 1
 ;;
esac
exit 0

chmod +x  /etc/init.d/haproxy # 可执行权限

创建haproxy.conf

vim /usr/local/haproxy/conf/haproxy.conf

 global
                log 127.0.0.1   local0
                log 127.0.0.1  local1 notice
                #log loghost    local0 info
                maxconn 50000
                chroot /usr/local/haproxy
                uid 99
                gid 99

                daemon
                nbproc 2
                pidfile /usr/local/haproxy/run/haproxy.pid
                #debug
                #quiet

 defaults
                log     global
                mode    tcp
                option  tcplog
                option  dontlognull
                option  forwardfor
                option  redispatch
                retries 3
                timeout connect     3000
                timeout client      50000
                timeout server 50000




frontend admin_stat
         bind *:8888
         mode http
         default_backend stats-back
frontend pxc-front
         bind    *:3306
         mode    tcp
         default_backend pxc-back
frontend pxc-onenode-front
         bind *:3308
         mode tcp
         default_backend pxc-onenode

backend pxc-back
        mode    tcp
        balance roundrobin
        option httpchk
        server node1 192.168.1.55:3306 check port 9200 inter 12000 rise 3 fall 3
        server node2 192.168.1.56:3307 check port 9200 inter 12000 rise 3 fall 3
        server node3 192.168.1.57:3307 check port 9200 inter 12000 rise 3 fall 3
backend pxc-onenode
        mode tcp
        balance roundrobin
        option httpchk
        server node1 192.168.1.55:3306 check port 9200 inter 12000 rise 3 fall 3
        server node2 192.168.1.56:3307 check port 9200 inter 12000 rise 3 fall 3  backup
        server node3 192.168.1.57:3307 check port 9200 inter 12000 rise 3 fall 3  backup
backend stats-back
        mode http
        balance roundrobin
        stats uri /admin?stats
        stats auth admin:admin
        stats realm Haproxy\ Statistics

Haproxy rsyslog 日志配置
vim /etc/rsyslog.conf 
#$ModLoad imudp      修改 $ModLoad imudp 
#$UDPServerRun 514  修改 $UDPServerRun 514
#添加 Haproxy  日志记录路径
local0.* /usr/local/haproxy/log/haproxy.log

重启rsyslog
systemctl restart  rsyslog.service
启动Haproxy 
service haproxy start
加入开机启动
chkconfig haproxy on
查看haproxy  是否启动
ps -ef | grep haproxy 
haproxy 访问端口8888 账号密码 admin:admin
http://192.168.1.56:8888/admin?stats
http://192.168.1.57:8888/admin?stats

七、编译安装keepalived

安装automake-1.15
tar -xvf /usr/local/src/automake-1.15.tar.gz -C /usr/local/src
cd /usr/local/src/automake-1.15
./configure && make && make install

编译keepalived

tar -xvf /usr/local/src/keepalived-1.3.2.tar.gz -C /usr/local/src
cd /usr/local/src/keepalived-1.3.2
./configure --prefix=/usr/local/keepalived
vmware 虚拟机安装 修改./lib/config.h
HAVE_DECL_RTA_ENCAP 1 修改为HAVE_DECL_RTA_ENCAP 0 如果不修改编译会报错
make && make install

拷贝keepalived 可执行文件到/usr/sbin/
cp /usr/local/keepalived/sbin/keepalived /usr/sbin/

创建keepalived 启动脚本

vim /etc/rc.d/init.d/keepalived

#!/bin/sh
#
# Startup script for the Keepalived daemon
#
# processname: keepalived
# pidfile: /var/run/keepalived.pid
# config: /etc/keepalived/keepalived.conf
# chkconfig: - 21 79
# description: Start and stop Keepalived

# Source function library
. /etc/rc.d/init.d/functions

# Source configuration file (we set KEEPALIVED_OPTIONS there)
. /etc/sysconfig/keepalived

RETVAL=0

prog="keepalived"

start() {
    echo -n $"Starting $prog: "
    daemon keepalived ${KEEPALIVED_OPTIONS}
    RETVAL=$?
    echo
    [ $RETVAL -eq 0 ] && touch /var/lock/subsys/$prog
}

stop() {
    echo -n $"Stopping $prog: "
    killproc keepalived
    RETVAL=$?
    echo
    [ $RETVAL -eq 0 ] && rm -f /var/lock/subsys/$prog
}

reload() {
    echo -n $"Reloading $prog: "
    killproc keepalived -1
    RETVAL=$?
    echo
}

# See how we were called.
case "$1" in
    start)
        start
        ;;
    stop)
        stop
        ;;
    reload)
        ;;
    restart)
        stop
        start
        ;;
    condrestart)
        if [ -f /var/lock/subsys/$prog ]; then
            stop
            start
        fi
        ;;
    status)
        status keepalived
        RETVAL=$?
        ;;
    *)
        echo "Usage: $0 {start|stop|reload|restart|condrestart|status}"
        RETVAL=1
esac

exit $RETVAL

chmod +x /etc/rc.d/init.d/keepalived #keepalived 可执行

配置keepalived 

cp /usr/local/keepalived/etc/sysconfig/keepalived  /etc/sysconfig/
创建/etc/keepalived 文件夹
mkdir -p /etc/keepalived

创建/etc/keepalived/keepalived.conf （192.168.1.56）

vim /etc/keepalived/keepalived.conf
! Configuration File for keepalived

global_defs {
   notification_email {
        admin@test.com    #发生故障时，接受信息的email地址
   }
   notification_email_from admin@test.com
   smtp_server 127.0.0.1
   smtp_connect_timeout 30
   router_id LVS_DEVEL
}

vrrp_script chk_haproxy {                           #自定义的监控脚本
    script "/etc/keepalived/chk_haproxy.sh"
    interval 2
    weight 2
}

vrrp_instance VI_1 {
    state BACKUP                                         #让VIP不抢回
    nopreempt
    interface eth0                                       #监听接口
    virtual_router_id 51
    priority 100                                            #优先级，backup机器上的优先级要小与这个值
    advert_int 1                                          #检查间隔
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    track_script {                                        #自定义的监控脚本
        chk_haproxy
    }
    virtual_ipaddress {                              #VIP地址，可以设多个
        192.168.4.58/21
    }
    notify_backup "/etc/init.d/haproxy restart"
    notify_fault "/etc/init.d/haproxy stop"
}

创建/etc/keepalived/keepalived.conf （192.168.1.57）

! Configuration File for keepalived

global_defs {
   notification_email {
        admin@test.com    #发生故障时，接受信息的email地址
   }
   notification_email_from admin@test.com
   smtp_server 127.0.0.1
   smtp_connect_timeout 30
   router_id LVS_DEVEL
}

vrrp_script chk_haproxy {                           #自定义的监控脚本
    script "/etc/keepalived/chk_haproxy.sh"
    interval 2
    weight 2
}

vrrp_instance VI_1 {
    state BACKUP                                         #让VIP不抢回
    nopreempt
    interface eth0                                       #监听接口
    virtual_router_id 51
    priority 99                                            #优先级，backup机器上的优先级要小与这个值
    advert_int 1                                          #检查间隔
    authentication {
        auth_type PASS
        auth_pass 1111
    }
    track_script {                                        #自定义的监控脚本
        chk_haproxy
    }
    virtual_ipaddress {                              #VIP地址，可以设多个
        192.168.4.58/21
    }
    notify_backup "/etc/init.d/haproxy restart"
    notify_fault "/etc/init.d/haproxy stop"
}

创建/etc/keepalived/chk_haproxy.sh

vim /etc/keepalived/chk_haproxy.sh

#!/bin/bash
if [ $(ps -C haproxy --no-header | wc -l) -eq 0 ];then
    /etc/init.d/haproxy start
fi
sleep 2
if [ $(ps -C haproxy --no-header | wc -l) -eq 0 ]; then
    /etc/init.d/keepalived stop
fi

chmod +x /etc/keepalived/chk_haproxy.sh #可执行权限

启动 keepalived

192.168.1.56
service keepalived start 

ip a
1: lo:  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
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
2: eth0:  mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:1b:50:a3 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.56/21 brd 192.168.7.255 scope global dynamic eth0
       valid_lft 44444sec preferred_lft 44444sec
    inet 192.168.4.58/21 scope global secondary eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fe1b:50a3/64 scope link 
       valid_lft forever preferred_lft forever

192.168.1.57
service keepalived start 
1: lo:  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
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
2: eth0:  mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:63:ae:85 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.57/21 brd 192.168.7.255 scope global dynamic eth0
       valid_lft 58142sec preferred_lft 58142sec
    inet6 fe80::20c:29ff:fe63:ae85/64 scope link 
       valid_lft forever preferred_lft forever

关闭 192.168.1.56 keepalived 进程

service keepalived stop
[root@56~]# ip a
1: lo:  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
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
2: eth0:  mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:1b:50:a3 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.56/21 brd 192.168.7.255 scope global dynamic eth0
       valid_lft 44258sec preferred_lft 44258sec
    inet6 fe80::20c:29ff:fe1b:50a3/64 scope link 
       valid_lft forever preferred_lft forever

查看 192.168.1.57

[root@57~]# ip a 
1: lo:  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
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
2: eth0:  mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:63:ae:85 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.57/21 brd 192.168.7.255 scope global dynamic eth0
       valid_lft 57869sec preferred_lft 57869sec
    inet 192.168.4.58/21 scope global secondary eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::20c:29ff:fe63:ae85/64 scope link 
       valid_lft forever preferred_lft forever

其它 keepalived  两个都配置为BACKUP 当第一个出现故障的时候恢复了不用抢占正在
提供服务的IP 服务不会闪断。 
    haproxy 配置双端口链接 3306 是同时负载3台服务器 3308 只是一台使用当第一台挂掉才会使用第二台。