采用Atlas+Keepalived实现MySQL读写分离、读负载均衡

原创作品，允许转载，转载时请务必以超链接形式标明文章 原始出处 、作者信息和本声明。否则将追究法律责任。http://sofar.blog.51cto.com/353572/1601552

==========================================================================================

一、基础介绍

==========================================================================================

1、背景描述

目前我们的高可用DB的代理层采用的是360开源的Atlas，从上线以来，已稳定运行2个多月。无论是从性能上，还是稳定性上，相比其他开源组件（amoeba、cobar、MaxScale、MySQL-Proxy等），还是很出色的。

当初我们之所以选择Atlas，主要看中它有以下优点：

(1)、基于mysql-proxy-0.8.2进行修改，代码完全开源；

(2)、比较轻量级，部署配置也比较简单；

(3)、支持DB读写分离；

(4)、支持从DB读负载均衡，并自动剔除故障从DB；

(5)、支持平滑上下线DB；

(6)、具备较好的安全机制（IP过滤、账号认证）；

(7)、版本更新、问题跟进、交流圈子都比较活跃。

在测试期间以及线上问题排查过程中，得到了360 Atlas作者朱超的热心解答，在此表示感谢。有关更多Atlas的介绍，我就不一一例举，可以参考以下链接：

https://github.com/Qihoo360/Atlas/blob/master/README_ZH.md

2、总体架构图

3、系统环境

CentOS 6.3 x86_64

==========================================================================================

二、安装部署

==========================================================================================

1、需注意的地方

(1)、本次安装不使用系统默认的glib库，之前的yum安装只是为了先解决依赖库的问题；

(2)、LUA库的版本不能太高，为5.1.x即可；

(3)、glib库的版本也不能太高，为glib-2.32.x即可；

(4)、对于编译不成功的情况，注意查看下面的说明。

2、GLIB依赖的基础库安装

# yum -y install *glib*

3、LUA库安装

http://ftp.gnu.org/pub/gnu/ncurses/ncurses-5.9.tar.gz

# tar xvzf ncurses-5.9.tar.gz

# cd ncurses-5.9

# ./configure --prefix=/usr/local

# make && make install

ftp://ftp.gnu.org/gnu/readline/readline-6.2.tar.gz

# tar xvzf readline-6.2.tar.gz

# cd readline-6.2

# ./configure --prefix=/usr/local

# make && make install

http://www.lua.org/ftp/lua-5.1.5.tar.gz

# tar xvzf lua-5.1.5.tar.gz

# cd lua-5.1.5

# make linux install

注意：

修改当前目录下的“Makefile”中的 INSTALL_TOP= /usr/local为 INSTALL_TOP= /usr/local/lua

主要是为了避免与系统自带的lua库发生冲突的可能

在“src/Makefile”文件中加入“-lncurses”，完整内容如下：

linux:

        $(MAKE) $(ALL) SYSCFLAGS="-DLUA_USE_LINUX" SYSLIBS="-Wl,-E -ldl -lncurses -lreadline"

4、GLIB库安装

ftp://sourceware.org/pub/libffi/libffi-3.0.13.tar.gz

# tar xvzf libffi-3.0.13.tar.gz

# cd libffi-3.0.13

# ./configure --prefix=/usr/local

# make && make install

http://ftp.gnu.org/pub/gnu/libiconv/libiconv-1.14.tar.gz

# tar xvzf libiconv-1.14.tar.gz

# cd libiconv-1.14

# ./configure --prefix=/usr/local

# make && make install

http://tukaani.org/xz/xz-5.0.5.tar.gz

# tar xvzf xz-5.0.5.tar.gz

# cd xz-5.0.5

# ./configure --prefix=/usr/local

# make && make install

# /sbin/ldconfig

http://ftp.gnome.org/pub/gnome/sources/glib/2.32/glib-2.32.4.tar.xz

# xz -d glib-2.32.4.tar.xz

# tar -xvf glib-2.32.4.tar

# cd glib-2.32.4

# ./configure --prefix=/usr/local/glib-2.32.4 \

--with-libiconv=/usr/local \

LIBFFI_CFLAGS="-I/usr/local/include" \

LIBFFI_LIBS="-L/usr/local/lib -lffi"

# make && make install

注意：编译报错处理

(1)、configure阶段

# vim ./glib/gconvert.c

注释掉第26、28行的内容

注释掉从61行到67行的内容

# vim ./configure

在7880行之上添加如下内容：

found_iconv=yes

(2)、make阶段

# ln -s /usr/local/lib/libffi-3.0.13/include/ffi.h /usr/local/include

# ln -s /usr/local/lib/libffi-3.0.13/include/ffitarget.h /usr/local/include

glib库需要安装在单独的目录“/usr/local/glib-2.32.4”，也是为了避免与系统自带的glib库发生冲突的可能

5、Atlas安装

(1)、其他基础组件安装

https://github.com/downloads/libevent/libevent/libevent-2.0.21-stable.tar.gz

# tar xvzf libevent-2.0.21-stable.tar.gz

# cd libevent-2.0.21-stable

# ./configure --prefix=/usr/local

# make && make install

http://www.openssl.org/source/openssl-1.0.1h.tar.gz

# tar xvzf openssl-1.0.1h.tar.gz

# cd openssl-1.0.1h

# ./config shared --prefix=/usr/local

# make && make install

(2)、MySQL安装（无需启动）

http://wwwNaNake.org/files/v2.8/cmake-2.8.10.2.tar.gz

# tar -xvzf cmake-2.8.10.2.tar.gz

# cd cmake-2.8.10.2

# ./bootstrap --prefix=/usr/local

# gmake --jobs=`grep processor /proc/cpuinfo | wc -l`

# gmake install

http://downloads.mysql.com/archives/get/file/mysql-5.5.24.tar.gz

# tar -xvzf mysql-5.5.24.tar.gz

# cd mysql-5.5.24

# rm-f CMakeCache.txt

# cmake -DCMAKE_INSTALL_PREFIX=/usr/local/mysql \

-DMYSQL_UNIX_ADDR=/var/run/mysql/mysql.sock \

-DDEFAULT_CHARSET=utf8 \

-DDEFAULT_COLLATION=utf8_general_ci \

-DEXTRA_CHARSETS=all \

-DWITH_MYISAM_STORAGE_ENGINE=1 \

-DWITH_INNOBASE_STORAGE_ENGINE=1 \

-DWITH_READLINE=1 \

-DENABLED_LOCAL_INFILE=1 \

-DWITH_EMBEDDED_SERVER=1 \

-DMYSQL_DATADIR=/data/dbdata/data \

-DMYSQL_TCP_PORT=3306

# make --jobs=`grep processor /proc/cpuinfo | wc -l`

# make install

(3)、DB中间件安装

https://github.com/Qihoo360/Atlas/archive/2.2.1.tar.gz

# tar xvzf Atlas-2.2.1.tar.gz

# cd Atlas-2.2.1

# ./configure --prefix=/usr/local/mysql-proxy \

--with-lua=/usr/local/lua \

--with-mysql=/usr/local/mysql \

GLIB_CFLAGS="-I/usr/local/glib-2.32.4/include/glib-2.0" \

GLIB_LIBS="-L/usr/local/glib-2.32.4/lib/glib-2.0 -lglib-2.0" \

GMODULE_CFLAGS="-I/usr/local/glib-2.32.4/include" \

GMODULE_LIBS="-L/usr/local/glib-2.32.4/lib -lgmodule-2.0" \

GTHREAD_CFLAGS="-I/usr/local/glib-2.32.4/include" \

GTHREAD_LIBS="-L/usr/local/glib-2.32.4/lib -lgthread-2.0" \

LUA_CFLAGS="-I/usr/local/lua/include" \

LUA_LIBS="-L/usr/local/lua/lib -llua-5.1" \

CFLAGS="-DHAVE_LUA_H -O2" \

LDFLAGS="-L/usr/local/lib -L/usr/local/lib64 -lm -ldl -lcrypto"

# make && make install

注意：

编译报错处理

# ln -s /usr/local/glib-2.32.4/lib/glib-2.0/include/glibconfig.h /usr/local/glib-2.32.4/include/glib-2.0

# cd /usr/local

# mv mysql-proxy atlas-2.2.1 && ln -s atlas-2.2.1 mysql-proxy

6、DB中间层配置

(1)、主配置

# vim /usr/local/mysql-proxy/conf/mysql-proxy.cnf

[mysql-proxy]
admin-username = sysadmin
admin-password = admin2356!@()

proxy-backend-addresses = 10.222.5.224:3306
proxy-read-only-backend-addresses = 10.240.95.107:3306,10.240.95.108:3306

pwds = health_check1:/iZxz+0GRoA=,health_check2:/iZxz+0GRoA=

daemon = true
keepalive = true

event-threads = 16

log-level = message
log-path = /usr/local/mysql-proxy/log
sql-log = ON

proxy-address = 0.0.0.0:3306
admin-address = 10.209.6.101:3307

charset = utf8

(2)、启动脚本

# vim /etc/init.d/mysql-proxy

#!/bin/sh
#
# mysql-proxy This script starts and stops the mysql-proxy daemon
#
# chkconfig: - 78 30
# processname: mysql-proxy
# description: mysql-proxy is a proxy daemon to mysql
# config: /usr/local/mysql-proxy/conf/mysql-proxy.cnf
# pidfile: /usr/local/mysql-proxy/log/mysql-proxy.pid
#
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin

DAEMON="/usr/local/mysql-proxy/bin/mysql-proxy"
CONFIGFILE="/usr/local/mysql-proxy/conf/mysql-proxy.cnf"
PIDFILE="/usr/local/mysql-proxy/log/mysql-proxy.pid"
LOCKFILE="/var/lock/subsys/mysql-proxy"
PROG=`basename $DAEMON`

RETVAL=0

start() {
    echo -n $"Starting ${PROG}......"
    [ -x $DAEMON ] || exit 5
    [ -f $CONFIGFILE ] || exit 6
    ${DAEMON} --defaults-file=${CONFIGFILE} || echo -n "${PROG} already running"

    RETVAL=$?
    echo
    [[ $RETVAL -eq 0 ]] && touch $LOCKFILE

    return $RETVAL
}

stop() {
    echo -n $"Stopping ${PROG}......"
    if [[ `ps aux | grep bin/mysql-proxy | grep -v grep | wc -l` -gt 0 ]]; then
        kill -TERM `ps -A -oppid,pid,cmd | grep bin/mysql-proxy | grep -v grep | awk '{print $2}'`
    fi

    RETVAL=$?
    echo
    [[ $RETVAL -eq 0 ]] && rm -f $LOCKFILE $PIDFILE

    return $RETVAL
}

restart() {
    stop
    sleep 1
    start
}

case "$1" in
start)
    start
    ;;

stop)
    stop
    ;;

restart)
    restart
    ;;

condrestart)
    [[ -e $LOCKFILE ]] && restart
    ;;

*)
    echo "Usage: $0 {start|stop|restart|condrestart}"
    RETVAL=1
    ;;
esac

exit $RETVAL

# chmod +x /etc/init.d/mysql-proxy

# chmod 0660 /usr/local/mysql-proxy/conf/mysql-proxy.cnf

# service mysql-proxy start

# ps aux | grep mysql-prox[y]

7、Atlas高可用【Keepalived】环境安装

http://rpm5.org/files/popt/popt-1.14.tar.gz

# tar xvzf popt-1.14.tar.gz

# cd popt-1.14

# ./configure --prefix=/usr/local

# make && make install

http://www.carisma.slowglass.com/~tgr/libnl/files/libnl-3.2.24.tar.gz

# tar xvzf libnl-3.2.24.tar.gz

# cd libnl-3.2.24

# ./configure --prefix=/usr/local

# make && make install

# ln -s /usr/local/include/libnl3/netlink /usr/local/include

# /sbin/ldconfig

http://www.keepalived.org/software/keepalived-1.2.10.tar.gz

# tar xvzf keepalived-1.2.10.tar.gz

# cd keepalived-1.2.10

# ./configure --prefix=/usr/local/keepalived

# make && make install

# ln -s /usr/local/keepalived/sbin/keepalived /usr/sbin

# ln -s /usr/local/keepalived/etc/sysconfig/keepalived /etc/sysconfig

# ln -s /usr/local/keepalived/etc/rc.d/init.d/keepalived /etc/init.d

# chkconfig --add keepalived

8、Atlas高可用【Keepalived】配置

# mkdir –p /etc/keepalived /data/scripts

(1)、主节点配置

# vim /etc/keepalived/keepalived.conf

global_defs {
    notification_email {
        [e-mail]
    }

    notification_email_from [e-mail]
    smtp_server 127.0.0.1
    smtp_connect_timeout 30
    router_id dbproxy1
}

vrrp_script chk_mysql_proxy_health {
    script "/data/scripts/keepalived_check_mysql_proxy.sh"
    interval 1
    weight -2
}

vrrp_instance VI_1 {
    state MASTER
    interface eth1
    virtual_router_id 51
    priority 100
    advert_int 1
    smtp_alert

    authentication {
        auth_type PASS
        auth_pass 123456
    }

    virtual_ipaddress {
        10.209.6.115
    }

    track_script {
        chk_mysql_proxy_health
    }

    notify_master "/data/scripts/notify.sh master"
    notify_bakcup "/data/scripts/notify.sh backup"
    notify_fault "/data/scripts/notify.sh fault"
}

(2)、备用节点配置

# vim /etc/keepalived/keepalived.conf

global_defs {
    notification_email {
        [e-mail]
    }

    notification_email_from [e-mail]
    smtp_server 127.0.0.1
    smtp_connect_timeout 30
    router_id dbproxy2
}

vrrp_script chk_mysql_proxy_health {
    script "/data/scripts/keepalived_check_mysql_proxy.sh"
    interval 1
    weight -2
}

vrrp_instance VI_1 {
    state BACKUP
    interface eth1
    virtual_router_id 51
    priority 90
    advert_int 1
    smtp_alert

    authentication {
        auth_type PASS
        auth_pass 123456
    }

    virtual_ipaddress {
        10.209.6.115
    }

    track_script {
        chk_mysql_proxy_health
    }

    notify_master "/data/scripts/notify.sh master"
    notify_bakcup "/data/scripts/notify.sh backup"
    notify_fault "/data/scripts/notify.sh fault"
}

(3)、VIP切换通知脚本

# vim /data/scripts/notify.sh

#!/bin/sh
PATH=/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/bin:/usr/local/sbin

KEEPALIVE_CONF="/etc/keepalived/keepalived.conf"

VIP=`grep -A 1 virtual_ipaddress ${KEEPALIVE_CONF} | tail -1 | sed 's/\t//g; s/ //g'`
ETH1_ADDR=`/sbin/ifconfig eth1 | awk '/inet addr:/{print $2}' | awk -F: '{print $2}'`

MONITOR="/usr/local/oms/agent/alarm/BusMonitorAgent"
TOKEN="ha_monitor"

function notify() {
    TITLE="$ETH1_ADDR to be $1: $VIP floating"
    CONTENT="vrrp transition, $ETH1_ADDR changed to be $1"
    ${MONITOR} -c 2 -f ${TOKEN} -t "${TITLE}" -i "${CONTENT}"
}

case "$1" in
master)
    notify master
    exit 0
    ;;

backup)
    notify backup
    exit 0
    ;;

fault)
    notify fault
    exit 0
    ;;

*)
    echo 'Usage: `basename $0` {master|backup|fault}'
    exit 1
    ;;
esac

(4)、DB中间层进程检查脚本

# vim /data/scripts/keepalived_check_mysql_proxy.sh

#!/bin/sh
PATH=/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/bin:/usr/local/sbin

if [[ `pgrep mysql-proxy | wc -l` -eq 0 ]]; then
    /sbin/service mysql-proxy start && sleep 5
    [[ -z `pgrep mysql-proxy` ]] && /sbin/service keepalived stop
fi

# chmod +x /data/scripts/*.sh

# service keepalived start

# ip addr show eth1

# ps aux | grep keepalive[d]

==========================================================================================

三、其他设置

==========================================================================================

1、Atlas服务监控

# vim /usr/local/mysql-proxy/bin/check_service.sh

#!/bin/sh
PATH=/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/bin:/usr/local/sbin

[[ $# -ne 3 ]] && echo "$0 端口号 协议类型 服务名" && exit 1

SRV_PORT=$1  ## 端口号
SRV_PROT=$2  ## 协议类型
SRV_NAME=$3  ## 服务名

MONITOR="/usr/local/oms/agent/alarm/BusMonitorAgent"
TOKEN="ha_monitor"

TITLE="${SRV_NAME}服务异常监控"
CONTENT="${SRV_NAME}服务发生异常，已自动拉起！"

## 是否已正确扫描
SCAN_FLAG=0

function RESTART_SRV_AND_ALERT()
{
    local CUR_SRV_NAME

    [[ $# -ne 1 ]] && exit 1
    CUR_SRV_NAME=$1

    TMP_SRV_NAME=`echo ${CUR_SRV_NAME} | tr '[A-Z]' '[a-z]'`
    [[ ! -f /etc/init.d/${TMP_SRV_NAME} ]] && TMP_SRV_NAME="${TMP_SRV_NAME}d"

    killall -9 ${TMP_SRV_NAME}

    if [[ -z `ps aux | grep ${TMP_SRV_NAME} | grep -v grep` ]]; then
        /sbin/service ${TMP_SRV_NAME} start >/dev/null 2>&1
    fi

    ${MONITOR} -c 2 -f ${TOKEN} -t "${TITLE}" -i "${CONTENT}"
    rm -f `pwd`/connect_error.log
}

ETH1_ADDR=`/sbin/ifconfig eth1 | awk -F ':' '/inet addr/{print $2}' | sed 's/[a-zA-Z ]//g'`
TMP_SRV_PROT=`echo ${SRV_PROT} | tr '[A-Z]' '[a-z]'`

if [[ "${TMP_SRV_PROT}" == "tcp" ]]; then
    PROT_OPT="S"
elif [[ "${TMP_SRV_PROT}" == "udp" ]]; then
    PROT_OPT="U"
else
    echo "未知的协议类型！" && exit 1
fi

## 最多扫描3次，成功一次即可，以避免网络抖动而导致误判
for ((i=0; i<3; i++)); do
    RETVAL=`/usr/bin/nmap -n -s${PROT_OPT} -p ${SRV_PORT} ${ETH1_ADDR} | grep open`
    [[ -n "${RETVAL}" ]] && SCAN_FLAG=1;break || sleep 10
done

## 1、针对Atlas服务端口不通的情况，也就是服务彻底挂掉
[[ ${SCAN_FLAG} -ne 1 ]] && RESTART_SRV_AND_ALERT ${SRV_NAME}

## 2、检查Atlas服务是否正常工作，也就是服务端口正常，但访问异常的情况【高权限DB用户】
mysqladmin -h${ETH1_ADDR} -uhealth_check1 -p123456 --connect-timeout=15 --shutdown-timeout=15 ping
[[ $? -ne 0 ]] && RESTART_SRV_AND_ALERT ${SRV_NAME}

## 3、检查Atlas服务是否正常工作，也就是服务端口正常，高权限DB用户访问也正常，但低权限
##    DB用户访问异常的情况【低权限DB用户】
mysqladmin -h${ETH1_ADDR} -uhealth_check2 -p123456 --connect-timeout=15 --shutdown-timeout=15 ping
[[ $? -ne 0 ]] && RESTART_SRV_AND_ALERT ${SRV_NAME}

2、Atlas访问日志切割

# vim /data/scripts/cut_and_clear_access_log.sh

#!/bin/sh
# 切割Atlas的访问日志，同时清理15天之前的日志
#
PATH=/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/bin:/usr/local/sbin

## mysql-proxy日志路径
LOGPATH="/usr/local/mysql-proxy/log"

[[ `/sbin/ip addr show eth1 | grep inet | wc -l` -eq 2 ]] || exit 1
cd ${LOGPATH}

## 日志切割
HISTORY_LOG_PATH=`date -d '-1 hour' +"%Y-%m-%d/sql_mysql-proxy_%H.log"`
[[ -d `dirname ${HISTORY_LOG_PATH}` ]] || mkdir -p `dirname ${HISTORY_LOG_PATH}`
cp -a sql_mysql-proxy.log ${HISTORY_LOG_PATH}

echo > sql_mysql-proxy.log

## 日志清理
HISTORY_LOG_PATH=`date -d '15 days ago' +'%Y-%m-%d'`
[[ -d ${HISTORY_LOG_PATH} ]] && rm -rf ${HISTORY_LOG_PATH}

3、crontab内容添加

# touch /var/lock/check_service.lock

# echo 'touch /var/lock/check_service.lock' >> /etc/rc.d/rc.local

# crontab -uroot -e

* * * * * (flock --timeout=0 /var/lock/check_service.lock /usr/local/mysql-proxy/bin/check_service.sh 3306 tcp mysql-proxy >/dev/null 2>&1)
00 * * * * /data/scripts/cut_and_clear_access_log.sh >/dev/null 2>&1

4、平滑设置功能

# mysql -h10.209.6.101 -P3307 -usysadmin -p'admin2356!@()'

本文出自 “人生理想在于坚持不懈” 博客，请务必保留此出处http://sofar.blog.51cto.com/353572/1601552