keepalived是一个是集群管理中保证集群高可用的一个服务软件,它具备心跳检测的功能。运用在redis上,则是具备主备切换的控制功能。
keepalived的主备切换方案的设计思想如下:
当 Master 与 Slave 均运作正常时, Master负责服务,Slave负责Standby;
当 Master 挂掉,Slave 正常时, Slave接管服务,同时关闭主从复制功能;
当 Master 恢复正常,则从Slave同步数据,同步数据之后关闭主从复制功能,恢复Master身份,于此同时Slave等待Master同步数据完成之后,恢复Slave身份。
然后依次循环。
下面就来大概讲下实现过程:
环境准备
主:10.15.34.21
备:10.15.34.22
VIP(虚拟IP):10.15.34.23,10.15.34.24
0.在主机和备机上分别创建用户部署redis和keepalived,如用户路径:/home/cachedb
1.在主机和备机上安装redis(过程略);
2.在主机和备机上安装keepalived(过程略);
3.修改主机和备机上的keepalived的配置文件keepalived.conf(可将原有的配置文件备份成keepalived.conf.bak)
1)主机的keepalived配置:
! Configuration File for keepalived global_defs { router_id LVS_REDIS_TEST_21 } vrrp_script chk_redis_1 { script "/home/cachedb/redis/scripts/redis_check.sh 10.15.34.21 38001" interval 10 timeout 2 fall 3 } vrrp_script chk_redis_2 { script "/home/cachedb/redis/scripts/redis_check.sh 10.15.34.21 38002" interval 10 timeout 2 fall 3 } vrrp_instance VI_1 { state BACKUP interface eth0 virtual_router_id 40 priority 150 nopreempt advert_int 1 authentication { auth_type PASS auth_pass 1111 } track_script { chk_redis_1 } virtual_ipaddress { 10.15.34.23 } notify_master "/home/cachedb/redis/scripts/redis_master.sh 10.15.34.21" notify_backup "/home/cachedb/redis/scripts/redis_backup.sh 10.15.34.21 38001 10.15.34.22 38001" notify_fault "/home/cachedb/redis/scripts/redis_fault.sh" notify_stop "/home/cachedb/redis/scripts/redis_stop.sh" } vrrp_instance VI_2 { state BACKUP interface eth0 virtual_router_id 41 priority 150 nopreempt advert_int 1 authentication { auth_type PASS auth_pass 1111 } track_script { chk_redis_2 } virtual_ipaddress { 10.15.34.54 } notify_master "/home/cachedb/redis/scripts/redis_master.sh 10.15.34.21 38002" notify_backup "/home/cachedb/redis/scripts/redis_backup.sh 10.15.34.21 38002 10.15.34.22 38002" notify_fault "/home/cachedb/redis/scripts/redis_fault.sh" notify_stop "/home/cachedb/redis/scripts/redis_stop.sh" }
2)备机的keepalived配置:
! Configuration File for keepalived global_defs { router_id LVS_REDIS_TEST } vrrp_script chk_redis_1 { script "/home/cachedb/redis/scripts/redis_check.sh 10.15.34.22 38001" interval 10 timeout 2 fall 3 } vrrp_script chk_redis_2 { script "/home/cachedb/redis/scripts/redis_check.sh 10.15.34.22 38002" interval 10 timeout 2 fall 3 } vrrp_instance VI_1 { state BACKUP interface eth0 virtual_router_id 40 priority 100 advert_int 1 authentication { auth_type PASS auth_pass 1111 } track_script { chk_redis_1 } virtual_ipaddress { 10.15.34.33 } notify_master "/home/cachedb/redis/scripts/redis_master.sh 10.15.34.22 38001" notify_backup "/home/cachedb/redis/scripts/redis_backup.sh 10.15.34.22 38001 10.15.34.21 38001" notify_fault "/home/cachedb/redis/scripts/redis_fault.sh" notify_stop "/home/cachedb/redis/scripts/redis_stop.sh" } vrrp_instance VI_2 { state BACKUP interface eth0 virtual_router_id 41 priority 100 advert_int 1 authentication { auth_type PASS auth_pass 1111 } track_script { chk_redis_2 } virtual_ipaddress { 10.15.34.34 } notify_master "/home/cachedb/redis/scripts/redis_master.sh 10.15.34.22 38002" notify_backup "/home/cachedb/redis/scripts/redis_backup.sh 10.15.34.22 38002 10.15.34.21 38002" notify_fault "/home/cachedb/redis/scripts/redis_fault.sh" notify_stop "/home/cachedb/redis/scripts/redis_stop.sh" }
注意:
a. 备机的keepalived配置的priority要比主机的低;
b. 备机不能有nopreempt属性;
c. virtual_router_id 这个主备配置的要对应一致,并且在局域网要唯一,注意使用默认的ID如51时要check改ID有木有被占用,若有就换一个。
d. interface eth0这个表名虚拟ip漂移的网卡为eth0,注意应该确保主备都有网卡eth0(没有的话也可以是eth1,eth2等)。可以用ifconfig来查看,并且主备都应该是相同的网卡eth~
配置完以后:
1)启动主机的redis服务,再在root下启动keepalived服务;
2)启动备机的redis服务,再在root下启动keepalived服务;
3)查看主机的角色:
./redis-cli -h 10.15.34.21 -p 38001
./redis-cli -h 10.15.34.22 -p 38001
查看主机的redis-state的log日志:(Log日志的位置由redis_*.sh脚本文件里指定)
4)尝试kill掉主机的redis服务,再在备机上查看role角色,得到:
可以看出它变成master了,而与此同时,它的redis-state.log日志如下:
1)我们不妨在备机上用root用户查看VIP的漂移情况:(虚拟IP的漂移需要一小段时间,不是立刻~)
ip a
1)而如果再重启之前master上的redis服务,会发现它的角色是slave了,这说明,master和salve发生了角色互换。
2)可以再尝试把备机上的redis服务kill掉,会发现主机上的角色会恢复成master;可以在主机上使用root用户来运行 ip a命令,会发现虚拟IP被漂移到主机上了。
系统日志位置:/var/log/messages