Redis主从自动failover

Redis主从架构持久化存在一个问题，即前次测试的结论，持久化需要配置在主实例上才能跨越实例保证数据不丢失，这样以来主实例在持久化数据到硬 盘的过程中，势必会造成磁盘的I/O等待，经过实际测试，这个持久化写硬盘的过程给应用程序带来的影响无法忍受；因而在大多数场景下，会考虑把持久化配置 在从实例上，当主实例宕机后，通过手动或者自动的方式将从实例提升为主实例，继续提供服务！当主实例恢复后，先从原从实例上同步数据，同步完成后再恢复到 原始的主从状态！要实现这种的要求，需要有keepalive的配合，一方面keepalive提供了VIP，可以避免修改应用程序连接，同时redis 实例的配置文件监听部分也需要修改为全网监听；另一方面keepalive定时调度脚本来监控主从实例的状态，根据具体情况进行切换！本文将重点介绍下使 用keepalive实现redis主从自动failover！

环境介绍
操作系统版本均为：rhel5.4 64bit
redis版本：2.6.4
redis实例端口均为：6379
redis实例密码均为：123
VIP：192.168.1.120
主实例为server11(192.168.1.112)
从实例为server12(192.168.1.113，开启快照持久化)

一：安装keepalive软件,server11安装完成后直接scp至server12上即可

    [root@server11 ~]# wget http://keepalived.org/software/keepalived-1.1.19.tar.gz
    [root@server11 ~]# tar -zxvf ../tarbag/keepalived-1.1.19.tar.gz
    [root@server11 ~]# cd keepalived-1.1.19/
    [root@server11 ~]# ./configure --prefix=/usr/local/keepalived && make && make install

二：配置主节点server11配置文件

    [root@server11 ~]# cat /usr/local/keepalived/etc/keepalived/keepalived.conf
    ! Configuration File for keepalived  

    global_defs {
     router_id LVS_DEVEL
    }  

    vrrp_script Monitor_redis {
     script "/usr/local/scripts/redis_monitor.sh"
     interval 2
     weight 2
    }  

    vrrp_instance VI_1{
     state MASTER
     interface eth0
     virtual_router_id 51
     mcast_src_ip 192.168.1.112
     priority  100
     advert_int 1
     authentication {
     auth_type PASS
     auth_pass password_123
    }
     track_script {
     Monitor_redis
    }
     virtual_ipaddress {
     192.168.1.120
     }
     notify_fault  /usr/local/scripts/redis_fault.sh
     notify_stop   /usr/local/scripts/redis_stop.sh    

    } 

三：配置从节点server12配置文件

    [root@server12 ~]# cat /usr/local/keepalived/etc/keepalived/keepalived.conf
    ! Configuration File for keepalived  

    global_defs {
     router_id LVS_DEVEL
    }  

    vrrp_script Monitor_redis {
     script "/usr/local/scripts/redis_monitor.sh"
     interval 2
     weight 2
    }  

    vrrp_instance VI_1{
     state BACKUP
     interface eth0
     virtual_router_id 51
     mcast_src_ip 192.168.1.113
     priority  99
     advert_int 1
     authentication {
     auth_type PASS
     auth_pass password_123
    }
     track_script {
     Monitor_redis
    }
     virtual_ipaddress {
     192.168.1.120
     }
     notify_master /usr/local/scripts/redis_master.sh
     notify_backup /usr/local/scripts/redis_backup.sh
     notify_fault  /usr/local/scripts/redis_fault.sh
     notify_stop   /usr/local/scripts/redis_stop.sh    

    } 

四：准备相关的脚本，主从实例上都需要存在这些脚本，同时注意脚本需要由可执行权限

    [root@server11 ~]# cat /usr/local/scripts/redis_monitor.sh
    #!/bin/bash
    ALIVE=$(/usr/local/redis2/bin/redis-cli -h 192.168.1.112 -p 6379 -a 123 PING)  

    if [ "$ALIVE" == "PONG" ]; then
        echo $ALIVE
        exit 0
        else
        echo $ALIVE
        killall -9 keepalived
        service network restart
        exit 1
    fi   

    [root@server11 ~]# sh /usr/local/scripts/redis_monitor.sh
    PONG  

    [root@server11 ~]# cat /usr/local/scripts/redis_master.sh
    #!/bin/bash
    REDISCLI="/usr/local/redis2/bin/redis-cli -h 192.168.1.112 -p 6379 -a 123"
    LOGFILE="/usr/local/redis2/var/keepalived-redis-state.log"   

    echo "[master]" >> $LOGFILE
    date >> $LOGFILE
    echo "Being master...." >> $LOGFILE 2>&1
    echo "Run SLAVEOF cmd ..." >> $LOGFILE
    $REDISCLI SLAVEOF 192.168.1.113 6379 >> $LOGFILE  2>&1
    sleep 10
    echo "Run SLAVEOF NO ONE cmd ..." >> $LOGFILE
    $REDISCLI SLAVEOF NO ONE >> $LOGFILE 2>&1   

    [root@server11 ~]# cat /usr/local/scripts/redis_backup.sh
    #!/bin/bash
    REDISCLI="/usr/local/redis2/bin/redis-cli -h 192.168.1.112 -p 6379 -a 123"
    LOGFILE="/usr/local/redis2/var/keepalived-redis-state.log"   

    echo "[backup]" >> $LOGFILE
    date >> $LOGFILE
    echo "Being slave...." >> $LOGFILE 2>&1
    sleep 15
    echo "Run SLAVEOF cmd ..." >> $LOGFILE
    $REDISCLI SLAVEOF 192.168.1.113 6379 >> $LOGFILE  2>&1   

    [root@server11 ~]# cat /usr/local/scripts/redis_stop.sh
    #!/bin/bash
    LOGFILE="/usr/local/redis2/var/keepalived-redis-state.log"
    echo "[stop]" >> $LOGFILE
    date >> $LOGFILE
    [root@server11 ~]# cat /usr/local/scripts/redis_fault.sh
    #!/bin/bash
    LOGFILE="/usr/local/redis2/var/keepalived-redis-state.log"
    echo "[fault]" >> $LOGFILE
    date >> $LOGFILE 

五：主从实例分别启动keepalive进程，测试VIP是否正常(这里就要修改redis配置文件的监听地址为0.0.0.0)

[root@server11 ~]# /usr/local/keepalived/sbin/keepalived -D -f  /usr/local/keepalived/etc/keepalived/keepalived.conf
[root@server11 ~]# tail -f /var/log/messages
Dec 12 09:25:49 server11 Keepalived_healthcheckers[7710]: Configuration is using : 5499 Bytes
Dec 12 09:25:49 server11 Keepalived_healthcheckers[7710]: Using LinkWatch kernel netlink reflector...
Dec 12 09:25:49 server11 Keepalived_vrrp[7712]: VRRP sockpool: [ifindex(2), proto(112), fd(12,13)]
Dec 12 09:25:49 server11 Keepalived_vrrp[7712]: VRRP_Script(Monitor_redis) succeeded
Dec 12 09:25:50 server11 Keepalived_vrrp[7712]: VRRP_Instance(VI_1{) Transition to MASTER STATE
Dec 12 09:25:51 server11 Keepalived_vrrp[7712]: VRRP_Instance(VI_1{) Entering MASTER STATE
Dec 12 09:25:51 server11 Keepalived_vrrp[7712]: VRRP_Instance(VI_1{) setting protocol VIPs.
Dec 12 09:25:51 server11 Keepalived_vrrp[7712]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120
Dec 12 09:25:51 server11 avahi-daemon[4519]: Registering new address record for 192.168.1.120 on eth0.
Dec 12 09:25:51 server11 Keepalived_healthcheckers[7710]: Netlink reflector reports IP 192.168.1.120 added
Dec 12 09:25:51 server11 Keepalived_vrrp[7712]: Netlink reflector reports IP 192.168.1.120 added
Dec 12 09:25:56 server11 Keepalived_vrrp[7712]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120  

[root@server11 ~]# ip a |grep 192
    inet 192.168.1.112/24 brd 192.168.1.255 scope global eth0
    inet 192.168.1.120/32 scope global eth0  

[root@server12 ~]# /usr/local/keepalived/sbin/keepalived -D -f /usr/local/keepalived/etc/keepalived/keepalived.conf
[root@server12 ~]# tail -f /var/log/messages
Dec 12 09:26:55 server12 Keepalived_healthcheckers[3106]: Configuration is using : 5595 Bytes
Dec 12 09:26:55 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Entering BACKUP STATE
Dec 12 09:26:55 server12 Keepalived_healthcheckers[3106]: Using LinkWatch kernel netlink reflector...
Dec 12 09:26:55 server12 Keepalived_vrrp[3108]: VRRP sockpool: [ifindex(2), proto(112), fd(12,13)]
Dec 12 09:26:55 server12 Keepalived_vrrp[3108]: VRRP_Script(Monitor_redis) succeeded  

[root@server11 ~]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |grep -A 3 'Replication'
# Replication
role:master
connected_slaves:1
slave0:192.168.1.113,6379,online

六：主实例写入测试数据，该脚本原则上会写入25条测试数据，不过由于未优化redis默认并发数，会导致一些写入请求失败，最终功写入231839条测 试数据，占内存总大小为25M左右，写入过程中可以观察主从实例的持久化文件变化情况，主实例的持久化文件维持在30k，从实例的则不断的扩展！

    [root@server11 ~]# cat test.sh
    #!/bin/bash
    REDISCLI="/usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 -n 1 SET"
    ID=1
    while(($ID<))
    do
    INSTANCE_NAME="i-2-$ID-VM"
    UUID=`cat /proc/sys/kernel/random/uuid`
    PRIVATE_IP_ADDRESS=10.`echo "$RANDOM % 255 + 1" | bc`.`echo "$RANDOM % 255 + 1" | bc`.`echo   

    "$RANDOM % 255 + 1" | bc`\
    CREATED=`date "+%Y-%m-%d %H:%M:%S"`
    $REDISCLI vm_instance:$ID:instance_name "$INSTANCE_NAME"
    $REDISCLI vm_instance:$ID:uuid "$UUID"
    $REDISCLI vm_instance:$ID:private_ip_address "$PRIVATE_IP_ADDRESS"
    $REDISCLI vm_instance:$ID:created "$CREATED"
    $REDISCLI vm_instance:$INSTANCE_NAME:id "$ID"
    ID=$(($ID+1))
    done   

    [root@server11 ~]# sh test.sh
    [root@server11 redis2]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |egrep   

    'used_memory_peak_human|db1:keys'
    used_memory_peak_human:24.98M
    db1:keys=231839,expires=0 

七：模拟主实例故障，观察日志输出，验证从实例是否能成功接管VIP，同时将实例变成读写模式

    [root@server11 ~]# killall -9  redis-server
    [root@server11 ~]# ip a |grep 192
        inet 192.168.1.112/24 brd 192.168.1.255 scope global eth0  

    [root@server11 ~]# ps -ef |grep redis
    root     15886  6458  0 09:49 pts/0    00:00:00 grep redis
    [root@server11 ~]# ps -ef |grep keep
    root     16029  6458  0 09:49 pts/0    00:00:00 grep keep  

    [root@server12 ~]# tail -f /usr/local/redis2/var/keepalived-redis-state.log
    [master]
    Wed Dec 12 09:48:52 CST 2012
    Being master....
    Run SLAVEOF cmd ...
    OK Already connected to specified master
    Run SLAVEOF NO ONE cmd ...
    OK  

    [root@server12 ~]# tail -f /var/log/messages
    Dec 12 09:48:51 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Transition to MASTER STATE
    Dec 12 09:48:52 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Entering MASTER STATE
    Dec 12 09:48:52 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) setting protocol VIPs.
    Dec 12 09:48:52 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120
    Dec 12 09:48:52 server12 Keepalived_vrrp[3108]: Netlink reflector reports IP 192.168.1.120 added
    Dec 12 09:48:52 server12 avahi-daemon[2921]: Registering new address record for 192.168.1.120 on eth0.
    Dec 12 09:48:52 server12 Keepalived_healthcheckers[3106]: Netlink reflector reports IP 192.168.1.120 added
    Dec 12 09:48:57 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120  

    [root@server12 ~]# ip a |grep 192
        inet 192.168.1.113/24 brd 192.168.1.255 scope global eth0
        inet 192.168.1.120/32 scope global eth0  

    [root@server12 ~]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |grep -A 3 'Replication'
    # Replication
    role:master
    connected_slaves:0 

    [root@server12 ~]# sh test.sh
    [root@server12 ~]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |egrep   

    'used_memory_peak_human|db1:keys'
    used_memory_peak_human:26.78M
    db1:keys=249925,expires=0 

九:主实例角色的恢复过程，使用shell脚本自动恢复

    [root@server11 ~]# ssh-keygen
    [root@server11 ~]# cd .ssh/
    [root@server11 .ssh]# ssh-copy-id -i id_rsa.pub root@192.168.1.113
    [root@server11 ~]# cat /usr/local/scripts/recover_mastart.sh
    #!/bin/sh
    ALIVE=$(/usr/local/redis2/bin/redis-cli -h 192.168.1.113 -p 6379 -a 123 PING)
    MDB=/usr/local/redis2/master_dump.rdb
    SDB=/usr/local/redis2/slave_dump.rdb  

    if [ "$ALIVE" == "PONG" ]; then
        echo $ALIVE
        scp root@192.168.1.113:$SDB  $MDB
        else
        echo $ALIVE
        exit 1
    fi   

    /usr/local/redis2/bin/redis-server /usr/local/redis2/etc/redis.conf
    /usr/local/keepalived/sbin/keepalived -D -f    

    /usr/local/keepalived/etc/keepalived/keepalived.conf  

    [root@server11 ~]# chmod +x  /usr/local/scripts/recover_mastart.sh
    [root@server11 ~]# sh /usr/local/scripts/recover_mastart.sh  

十：验证数据完整性和主从角色恢复情况

    [root@server11 ~]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |egrep 'used_memory_peak_human|db1:keys'
    used_memory_peak_human:26.78M
    db1:keys=249925,expires=0 

    [root@server11 ~]#  /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |grep -A 3 'Replication'
    # Replication
    role:master
    connected_slaves:1
    slave0:192.168.1.113,6379,online  

    [root@server12 ~]#  /usr/local/redis2/bin/redis-cli -h 192.168.1.113 -a 123 info |grep -A 3 'Replication'
    # Replication
    role:slave
    master_host:192.168.1.112
    master_port:6379  

    [root@server12 ~]# /usr/local/redis2/bin/redis-cli -h 192.168.1.120 -a 123 info |egrep 'used_memory_peak_human|db1:keys'
    used_memory_peak_human:26.78M
    db1:keys=249925,expires=0 

    主实例keepalive日志：
    [root@server11 ~]# tail -f /var/log/messages
    Dec 12 10:08:13 server11 Keepalived_vrrp[20231]: VRRP sockpool: [ifindex(2), proto(112), fd(11,12)]
    Dec 12 10:08:13 server11 Keepalived_vrrp[20231]: VRRP_Script(Monitor_redis) succeeded
    Dec 12 10:08:13 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Transition to MASTER STATE
    Dec 12 10:08:13 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Received higher prio advert
    Dec 12 10:08:13 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Entering BACKUP STATE
    Dec 12 10:08:15 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) forcing a new MASTER election
    Dec 12 10:08:16 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Transition to MASTER STATE
    Dec 12 10:08:17 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Entering MASTER STATE
    Dec 12 10:08:17 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) setting protocol VIPs.
    Dec 12 10:08:17 server11 Keepalived_healthcheckers[20230]: Netlink reflector reports IP 192.168.1.120 added
    Dec 12 10:08:17 server11 Keepalived_vrrp[20231]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120
    Dec 12 10:08:17 server11 Keepalived_vrrp[20231]: Netlink reflector reports IP 192.168.1.120 added
    Dec 12 10:08:17 server11 avahi-daemon[4519]: Registering new address record for 192.168.1.120 on eth0.  

    [root@server11 ~]# ip a |grep 192
        inet 192.168.1.112/24 brd 192.168.1.255 scope global eth0
        inet 192.168.1.120/32 scope global eth0  

    从实例keepalive日志：
    [root@server12 ~]# tail -f /var/log/messages
    Dec 12 09:56:01 server12 last message repeated 4 times
    Dec 12 10:08:13 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Received lower prio advert, forcing new election
    Dec 12 10:08:13 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Sending gratuitous ARPs on eth0 for 192.168.1.120
    Dec 12 10:08:15 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Received higher prio advert
    Dec 12 10:08:15 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) Entering BACKUP STATE
    Dec 12 10:08:15 server12 Keepalived_vrrp[3108]: VRRP_Instance(VI_1{) removing protocol VIPs.
    Dec 12 10:08:15 server12 Keepalived_healthcheckers[3106]: Netlink reflector reports IP 192.168.1.120 removed
    Dec 12 10:08:15 server12 Keepalived_vrrp[3108]: Netlink reflector reports IP 192.168.1.120 removed
    Dec 12 10:08:15 server12 avahi-daemon[2921]: Withdrawing address record for 192.168.1.120 on eth0.  

    从实例角色转换日志：
    [root@server12 ~]# tail -f /usr/local/redis2/var/keepalived-redis-state.log
    [backup]
    Wed Dec 12 10:08:15 CST 2012
    Being slave....
    Run SLAVEOF cmd ...
    OK