資訊專欄INFORMATION COLUMN
摘要：雙主是一個比較簡單的高可用架構(gòu)，適用于中小集群，今天就說說怎么用做的高可用。缺點也比較明顯，就是增加從節(jié)點的情況下，從節(jié)點不會主動切換同步對象，而且腳本需要自己實現(xiàn)，有一定風(fēng)險。
雙主 + keepalived 是一個比較簡單的 MySQL 高可用架構(gòu)，適用于中小 MySQL 集群，今天就說說怎么用 keepalived 做 MySQL 的高可用。
1 概述 1.1 keepalived 簡介簡單地說，keepalived 就是通過管理 VIP 來實現(xiàn)機器的高可用的，在使用 keepalived 的情況下，只有一臺服務(wù)器能夠提供服務(wù)(通過 VIP 來實現(xiàn))，當(dāng) Master 主機宕機后，VIP 會自動飄移到另一臺服務(wù)器
keepalived 采用 Master/Slave 模式, 在 Master 上設(shè)置配置文件的 VIP，當(dāng) Master 宕機后，VIP 自動漂移到另一臺 keepalived 服務(wù)器上
keepalived 可以用來做各種軟件的高可用集群，它會一直檢測服務(wù)器的狀態(tài)，如果有一臺服務(wù)器宕機，或工作出現(xiàn)故障，keepalived 將檢測到，并將有故障的服務(wù)器從系統(tǒng)中剔除，同時使用其他服務(wù)器代替該服務(wù)器的工作，當(dāng)服務(wù)器工作正常后 keepalived 自動將服務(wù)器加入到服務(wù)器群中。
1.2 keepalived 配合雙主keepalived 使用默認(rèn)配置只能做到主機級別的高可用，但是我們的 MySQL 要做高可用至少要增加以下功能
能夠檢測 MySQL 服務(wù)狀態(tài)
主節(jié)點 read_only=0，備節(jié)點 read_only=1
切換時，備節(jié)點要等待主節(jié)點同步完成
所以，keepalived 實現(xiàn) MySQL 高可用需要使用自定義腳本來進行擴展
2 環(huán)境準(zhǔn)備 2.1 數(shù)據(jù)庫環(huán)境操作前已經(jīng)準(zhǔn)備好了一套主主架構(gòu)數(shù)據(jù)庫，搭建方法參考 MySQL集群搭建(2)-主主從模式
節(jié)點信息| IP | 系統(tǒng) | 端口 | MySQL版本 | 節(jié)點 | 讀寫 | 說明 |
|---|---|---|---|---|---|---|
| 10.0.0.247 | Centos6.5 | 3306 | 5.7.9 | Master | 讀寫 | 主節(jié)點 |
| 10.0.0.248 | Centos6.5 | 3306 | 5.7.9 | Standby | 只讀,可切換為讀寫 | 備主節(jié)點 |
| 簡稱 | VIP | 類型 |
|---|---|---|
| RW-VIP | 10.0.0.237 | 讀寫VIP |
[client] port = 3306 default-character-set=utf8mb4 socket = /data/mysql_db/test_db/mysql.sock [mysqld] datadir = /data/mysql_db/test_db basedir = /usr/local/mysql57 tmpdir = /tmp socket = /data/mysql_db/test_db/mysql.sock pid-file = /data/mysql_db/test_db/mysql.pid skip-external-locking = 1 skip-name-resolve = 1 port = 3306 server_id = 2473306 default-storage-engine = InnoDB character-set-server = utf8mb4 default_password_lifetime=0 auto_increment_offset = 1 auto_increment_increment = 2 #### log #### log_timestamps=system log_bin = /data/mysql_log/test_db/mysql-bin log_bin_index = /data/mysql_log/test_db/mysql-bin.index binlog_format = row relay_log_recovery=ON relay_log=/data/mysql_log/test_db/mysql-relay-bin relay_log_index=/data/mysql_log/test_db/mysql-relay-bin.index log_error = /data/mysql_log/test_db/mysql-error.log #### replication #### log_slave_updates = 1 replicate_wild_ignore_table = information_schema.%,performance_schema.%,sys.% #### semi sync replication settings ##### plugin_dir=/usr/local/mysql57/lib/plugin plugin_load = "rpl_semi_sync_master=semisync_master.so;rpl_semi_sync_slave=semisync_slave.so" loose_rpl_semi_sync_master_enabled = 1 loose_rpl_semi_sync_slave_enabled = 1Slave 參考配置
[client] port = 3306 default-character-set=utf8mb4 socket = /data/mysql_db/test_db/mysql.sock [mysqld] datadir = /data/mysql_db/test_db basedir = /usr/local/mysql57 tmpdir = /tmp socket = /data/mysql_db/test_db/mysql.sock pid-file = /data/mysql_db/test_db/mysql.pid skip-external-locking = 1 skip-name-resolve = 1 port = 3306 server_id = 2483306 default-storage-engine = InnoDB character-set-server = utf8mb4 default_password_lifetime=0 auto_increment_offset = 2 auto_increment_increment = 2 #### log #### log_timestamps=system log_bin = /data/mysql_log/test_db/mysql-bin log_bin_index = /data/mysql_log/test_db/mysql-bin.index binlog_format = row relay_log_recovery=ON relay_log=/data/mysql_log/test_db/mysql-relay-bin relay_log_index=/data/mysql_log/test_db/mysql-relay-bin.index log_error = /data/mysql_log/test_db/mysql-error.log #### replication #### log_slave_updates = 1 replicate_wild_ignore_table = information_schema.%,performance_schema.%,sys.% #### semi sync replication settings ##### plugin_dir=/usr/local/mysql57/lib/plugin plugin_load = "rpl_semi_sync_master=semisync_master.so;rpl_semi_sync_slave=semisync_slave.so" loose_rpl_semi_sync_master_enabled = 1 loose_rpl_semi_sync_slave_enabled = 12.2 創(chuàng)建監(jiān)控用的賬號
- 由于是測試環(huán)境，賬號密碼設(shè)置比較隨便 create user monitor@"localhost" identified by "monitor"; grant all on *.* to monitor@"localhost"; flush privileges;2.3 安裝 keepalived
我們在 Master 和 Slave 上部署 keepalived
1). yum 安裝如果有對應(yīng)的 yum 源，直接安裝就可以了
yum install -y keepalived2). 源碼安裝
下載安裝包, 下載地址 keepalived, 使用 1.2.24 版本舉例
# 安裝依賴 yum install -y gcc popt-devel openssl openssl-devel libssl-dev libnl-devel popt-devel libnfnetlink-devel # 下載包 wget http://www.keepalived.org/software/keepalived-1.2.24.tar.gz # 解壓安裝 tar -xvz -f keepalived-1.2.24.tar.gz cd keepalived-1.2.24 ./configure --prefix=/usr/local/keepalived make && make install cp /usr/local/keepalived/sbin/keepalived /usr/sbin/ cp /usr/local/keepalived/etc/rc.d/init.d/keepalived /etc/init.d/ cp /usr/local/keepalived/etc/sysconfig/keepalived /etc/sysconfig/ mkdir /etc/keepalived/ cp /usr/local/keepalived/etc/keepalived/keepalived.conf /etc/keepalived/3 配置高可用 3.1 keepalived 配置
打開 /etc/keepalived/keepalived.conf 文件, 按照實際情況加上下面的配置
global_defs {
router_id MYSQL_MM # 標(biāo)識
vrrp_skip_check_adv_addr
vrrp_strict # 嚴(yán)格執(zhí)行 VRRP 協(xié)議規(guī)范
vrrp_garp_interval 0
vrrp_gna_interval 0
}
vrrp_script check_mysql {
script "/bin/sh /etc/keepalived/keepalived_mysql_check.sh" # 檢查腳本
interval 10 # 檢查周期
}
vrrp_instance MYSQL_MM {
state BACKUP # 都設(shè)為 BACKUP，避免起來后搶占
interface eth0 # 網(wǎng)卡名稱，根據(jù)實際情況填寫
virtual_router_id 243 # 用來區(qū)分 VRRP 組播的標(biāo)記，取值 0-255
priority 100
advert_int 1
nopreempt # 設(shè)為非搶占
authentication {
auth_type PASS
auth_pass 1111
}
# Master 節(jié)點可以注釋掉下面語句，防止啟動 keepalived 的時候執(zhí)行腳本
notify_master "/bin/sh /etc/keepalived/keepalived_mysql_start.sh" # 變?yōu)?MASTER 時執(zhí)行
virtual_ipaddress {
10.0.0.237
}
# Slave 節(jié)點可以注釋下面檢查腳本，Slave 沒有必要一直檢查
track_script {
check_mysql
}
}
3.2 配置檢查腳本
打開 /etc/keepalived/keepalived_mysql_check.sh， 寫入檢測腳本
#!/bin/sh
# @Author: chengqm
# MySQL 檢測腳本
MyPath=$(cd $(dirname $0); pwd)
cd $MyPath
ThisTime=`date "+%F %T"`
log_file="/var/log/keepalived_mysql.log"
# MySQL 連接方式，根據(jù)實際情況調(diào)整
export MYSQL_PWD="monitor"
MYSQL_USER="monitor"
MYSQL_SOCKET="/data/mysql_db/test_db/mysql.sock"
mysql_connect="mysql -u${MYSQL_USER} -S${MYSQL_SOCKET} "
# 美化輸出
function techo() {
message=$1
message_level=$2
if [ -e $message_level ];then
message_level="info"
fi
echo "`date "+%F %T"` - [${message_level}] $message" >> $log_file
}
# 檢查函數(shù), 正常返回 0
function check {
ret=`$mysql_connect -N -e "select 1 as value"`
if [ $? -ne 0 ] || [ $ret -ne "1" ];then
return 1
else
return 0
fi
}
function read_only {
param=$1
$mysql_connect -e "set global read_only = ${param}"
techo "設(shè)置是否只讀 read_only ${param}"
}
# 失效轉(zhuǎn)移
function failover {
techo "開始執(zhí)行失效轉(zhuǎn)移"
# 1. 停止 keepalived
killall keepalived
# 2. 如果還能執(zhí)行的話，設(shè)為 read_only
read_only 1
if [ $? -eq 0 ];then
# 3. 如果還能執(zhí)行，kill 所有的連接
$mysql_connect -e "select concat("KILL ",id,";") from information_schema.processlist where user!="root" AND db is not null into outfile "/tmp/kill.txt.${ThisTime}";"
if [ $? -eq 0 ];then
$mysql_connect -e "source /tmp/kill.txt.${ThisTime};"
fi
fi
# 4. 其他操作，比如說自動關(guān)機
techo "失效轉(zhuǎn)移執(zhí)行成功，當(dāng)前數(shù)據(jù)庫關(guān)閉訪問"
}
# 有問題檢查 4 次
for ((i=1; i<=4; i ++))
do
check
if [ $? -eq 0 ];then
techo "MySQL is ok"
# 正常退出腳本
exit 0
else
techo "Connection failed $i time(s)"
sleep 1
fi
done
techo "無法連接當(dāng)前數(shù)據(jù)庫"
# 失效轉(zhuǎn)移
failover
注意：腳本沒有經(jīng)過嚴(yán)格測試，需要根據(jù)實際情況調(diào)整
3.3 配置提升為 Master 時執(zhí)行的腳本打開 /bin/sh /etc/keepalived/keepalived_mysql_start.sh", 寫入腳本內(nèi)容
#!/bin/sh
# @Author: chengqm
# keepalived 變?yōu)?Master 時執(zhí)行
MyPath=$(cd $(dirname $0); pwd)
cd $MyPath
ThisTime=`date "+%F %T"`
log_file="/var/log/keepalived_mysql.log"
# MySQL 連接方式，根據(jù)實際情況調(diào)整
export MYSQL_PWD="monitor"
MYSQL_USER="monitor"
MYSQL_SOCKET="/data/mysql_db/test_db/mysql.sock"
mysql_connect="mysql -u${MYSQL_USER} -S${MYSQL_SOCKET} "
# 美化輸出
function techo() {
message=$1
message_level=$2
if [ -e $message_level ];then
message_level="info"
fi
echo "`date "+%F %T"` - [${message_level}] $message" >> $log_file
}
# 檢查函數(shù), 正常返回 0
function check {
ret=`$mysql_connect -N -e "select 1 as value"`
if [ $? -ne 0 ] || [ $ret -ne "1" ];then
return 1
else
return 0
fi
}
# 獲取 slave status 的信息
function slave_info() {
tmp_file=/tmp/slave_info.tmp
$mysql_connect -e "show slave statusG" > /tmp/slave_info.tmp
slave_sql=`grep "Slave_SQL_Running:" $tmp_file | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
seconds_behind_master=`grep "Seconds_Behind_Master:" $tmp_file | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
master_log_file=`grep "Master_Log_File:" $tmp_file | head -1 | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
master_log_pos=`grep "Read_Master_Log_Pos:" $tmp_file | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
relay_master_log_file=`grep "Relay_Master_Log_File:" $tmp_file | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
exec_master_log_pos=`grep "Exec_Master_Log_Pos:" $tmp_file | sed "s/s*//g" | tr "A-Z" "a-z" | awk -F":" "{print $2}"`
}
# 設(shè)置是否可讀
function read_only {
param=$1
$mysql_connect -e "set global read_only = ${param}"
techo "設(shè)置是否只讀 read_only ${param}"
}
# 處理數(shù)據(jù)同步
function sync_master_log() {
# 如果是數(shù)據(jù)一致性優(yōu)先，等待同步完畢。如果是服務(wù)可用性優(yōu)先，可以注銷下面的代碼
slave_info
if [ $slave_sql == "yes" ];then
techo "當(dāng)前同步位置 Master ${master_log_file} ${master_log_pos}"
techo "等待同步到 Master ${master_log_file} ${master_log_pos}"
$mysql_connect -e "select master_pos_wait("$master_log_file", $master_log_pos);" > /dev/null
techo "同步完畢"
fi
}
techo "當(dāng)前數(shù)據(jù)庫提升為主庫"
check
if [ $? -ne 0 ];then
techo "無法連接當(dāng)前數(shù)據(jù)庫"
exit 1
fi
# 等待同步
sync_master_log
# 設(shè)為可寫
read_only 0
注意：腳本沒有經(jīng)過嚴(yán)格測試，需要根據(jù)實際情況調(diào)整
3.4 啟動 keepalived由于配置了 BACKUP 模式，所以兩個 keepalived 先起來的是主，先后在主備節(jié)點執(zhí)行
/etc/init.d/keepalived start
檢查 /var/log/message 日志，確認(rèn) keepalived 沒有報錯
檢查 Master IP 狀態(tài), 確認(rèn)設(shè)置了 VIP
