redis高可用方案之sentinel(哨兵集群)
今天给大家介绍下redis高可用方案,redis自带的sentinel,简单的口语描述下什么是sentinel(哨兵集群),哨兵集群能够将redis主从封装起来,实现redis主宕机,redis备将接管主,再次恢复的时候自动升级为从。网上有很多介绍redis+vip的方法,这个很简单就是当主宕机之后,实现vip的漂移。这种方法确实是可以的,但是没有办法实现主从资源利用最大化,我要给大家介绍的是如何使用sentinel+redis+python(宕机报警以后在添加)实现redis高可用以及redis读写分离。
1.安装redis(我这边使用了2个,生产上建议是用3个)
具体安装redis就不介绍了。我这是使用的是ansible安装
我这里安装的是2个,192.168.7.30准备用来的做主redis用来写数据,192.168.7.31是从用来读数据。
2.配置redis
因为使用的是ansible,所以配置文件是默认配置好的,不用做修改,可以直接启动就行了。
172.16.10.30主上的配置
[root@redis-master etc]# cat redis_16399.conf bind 172.16.10.30protected-mode yesport 16399tcp-backlog 511timeout 0tcp-keepalive 300daemonize yessupervised nopidfile /usr/local/redis/run/redis_16399.pidloglevel noticelogfile "/usr/local/redis/logs/redis_16399.log"databases 16save 900 1save 300 10save 60 10000stop-writes-on-bgsave-error yesrdbcompression yesrdbchecksum yesdbfilename dump.rdbdir /data/redis/16399slave-serve-stale-data yesslave-read-only yesrepl-diskless-sync norepl-diskless-sync-delay 5repl-disable-tcp-nodelay noslave-priority 100appendonly yesappendfilename "appendonly.aof"appendfsync everysecno-appendfsync-on-rewrite noauto-aof-rewrite-percentage 100auto-aof-rewrite-min-size 64mbaof-load-truncated yeslua-time-limit 5000slowlog-log-slower-than 10000slowlog-max-len 128latency-monitor-threshold 0notify-keyspace-events ""hash-max-ziplist-entries 512hash-max-ziplist-value 64list-max-ziplist-size -2list-compress-depth 0set-max-intset-entries 512zset-max-ziplist-entries 128zset-max-ziplist-value 64hll-sparse-max-bytes 3000activerehashing yesclient-output-buffer-limit normal 0 0 0client-output-buffer-limit slave 256mb 64mb 60client-output-buffer-limit pubsub 32mb 8mb 60hz 10aof-rewrite-incremental-fsync yes[root@redis-master etc]# /etc/init.d/redis_16399 start Starting Redis server...[root@redis-master etc]# netstat -nltp |grep 16399 tcp 0 0 172.16.10.30:16399 0.0.0.0:* LISTEN 5521/redis-server 1 [root@redis-master etc]#
172.16.10.31上的配置
[root@master-slave etc]# grep -rn slaveof redis_16399.conf 16:slaveof 172.16.10.30 16399 #从配置文件比主多这行[root@master-slave etc]# /etc/init.d/redis_16399 start Starting Redis server...[root@master-slave etc]# netstat -nltpa |grep 16399 tcp 0 0 172.16.10.31:16399 0.0.0.0:* LISTEN 10466/redis-server tcp 0 0 172.16.10.31:18936 172.16.10.30:16399 ESTABLISHED 10466/redis-server [root@master-slave etc]#
从上面就可以看出主从已经配置完成。
3.配置sentinel
sentinel的配置文件所有的节点基本是一样的,可以部署在不是redis的节点上,进行更加全面的监控
[root@redis-master etc]# cat sentinel_5001.conf port 5001pidfile "/usr/local/redis/run/sentinel_5001.pid"loglevel noticelogfile "/usr/local/redis/logs/sentinel_5001.log"daemonize yesprotected-mode nosentinel monitor mymaster 172.16.10.30 16399 2sentinel down-after-milliseconds mymaster 5000sentinel failover-timeout mymaster 60000[root@redis-master etc]#
4.启动sentinel
172.16.10.30启动sentinel
[root@master-slave etc]# cat sentnel_5001.conf port 5001pidfile "/usr/local/redis/run/sentinel_5001.pid"loglevel noticelogfile "/usr/local/redis/logs/sentinel_5001.log"daemonize yesprotected-mode nosentinel monitor mymaster 172.16.10.30 16399 2sentinel down-after-milliseconds mymaster 5000sentinel failover-timeout mymaster 60000[root@master-slave etc]# [root@redis-master etc]# /usr/local/redis/bin/redis-sentinel /usr/local/redis/etc/sentinel_5001.conf [root@redis-master etc]# tail -30f /usr/local/redis/logs/sentinel_5001.log 5704:X 08 Dec 22:46:44.977 # Not listening to IPv6: unsupproted _._ _.-``__ ''-._ _.-`` `. `_. ''-._ Redis 3.2.11 (00000000/0) 64 bit .-`` .-```. ```\/ _.,_ ''-._ ( ' , .-` | `, ) Running in sentinel mode |`-._`-...-` __...-.``-._|'` _.-'| Port: 5001 | `-._ `._ / _.-' | PID: 5704 `-._ `-._ `-./ _.-' _.-' |`-._`-._ `-.__.-' _.-'_.-'| | `-._`-._ _.-'_.-' | http://redis.io `-._ `-._`-.__.-'_.-' _.-' |`-._`-._ `-.__.-' _.-'_.-'| | `-._`-._ _.-'_.-' | `-._ `-._`-.__.-'_.-' _.-' `-._ `-.__.-' _.-' `-._ _.-' `-.__.-' 5704:X 08 Dec 22:46:44.986 # Sentinel ID is 453837ef09fc1a41332df32487882f2e422337625704:X 08 Dec 22:46:44.986 # +monitor master mymaster 172.16.10.30 16399 quorum 25704:X 08 Dec 22:46:44.987 * +slave slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 16399
到这里本来就已经监控起来了,上面我们配置只有2个节点的时候才会发生切换,我们现在将的172.16.10.31上的sentinel启动起来。
[root@master-slave etc]# vim sentnel_5001.conf [root@master-slave etc]# /usr/local/redis/bin/redis-sentinel /usr/local/redis/etc/sentnel_5001.conf [root@master-slave etc]# tail -30f /usr/local/redis/logs/sentinel_5001.log 10615:X 08 Dec 22:51:05.232 # Not listening to IPv6: unsupproted _._ _.-``__ ''-._ _.-`` `. `_. ''-._ Redis 3.2.11 (00000000/0) 64 bit .-`` .-```. ```\/ _.,_ ''-._ ( ' , .-` | `, ) Running in sentinel mode |`-._`-...-` __...-.``-._|'` _.-'| Port: 5001 | `-._ `._ / _.-' | PID: 10615 `-._ `-._ `-./ _.-' _.-' |`-._`-._ `-.__.-' _.-'_.-'| | `-._`-._ _.-'_.-' | http://redis.io `-._ `-._`-.__.-'_.-' _.-' |`-._`-._ `-.__.-' _.-'_.-'| | `-._`-._ _.-'_.-' | `-._ `-._`-.__.-'_.-' _.-' `-._ `-.__.-' _.-' `-._ _.-' `-.__.-' 10615:X 08 Dec 22:51:05.234 # Sentinel ID is 0ac7df95b3e057a0ccb4a10f1cc8cc91e071526e10615:X 08 Dec 22:51:05.234 # +monitor master mymaster 172.16.10.30 16399 quorum 210615:X 08 Dec 22:51:05.235 * +slave slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 1639910615:X 08 Dec 22:51:06.697 * +sentinel sentinel 453837ef09fc1a41332df32487882f2e42233762 172.16.10.30 5001 @ mymaster 172.16.10.30 16399
到这里我们已经完成了sentinel集群的搭建
5.测试
测试主从
测试sentinel
停掉 172.16.10.30的redis
[root@redis-master init.d]# /etc/init.d/redis_16399 stop Stopping ...Redis stopped
sentinel fallover日志
10615:X 08 Dec 22:52:08.497 # +sdown master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.553 # +odown master mymaster 172.16.10.30 16399 #quorum 2/210615:X 08 Dec 22:52:08.553 # +new-epoch 110615:X 08 Dec 22:52:08.553 # +try-failover master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.554 # +vote-for-leader 0ac7df95b3e057a0ccb4a10f1cc8cc91e071526e 110615:X 08 Dec 22:52:08.556 # 453837ef09fc1a41332df32487882f2e42233762 voted for 0ac7df95b3e057a0ccb4a10f1cc8cc91e071526e 110615:X 08 Dec 22:52:08.655 # +elected-leader master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.655 # +failover-state-select-slave master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.710 # +selected-slave slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.710 * +failover-state-send-slaveof-noone slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:08.782 * +failover-state-wait-promotion slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:09.715 # +promoted-slave slave 172.16.10.31:16399 172.16.10.31 16399 @ mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:09.715 # +failover-state-reconf-slaves master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:09.785 # +failover-end master mymaster 172.16.10.30 1639910615:X 08 Dec 22:52:09.785 # +switch-master mymaster 172.16.10.30 16399 172.16.10.31 1639910615:X 08 Dec 22:52:09.785 * +slave slave 172.16.10.30:16399 172.16.10.30 16399 @ mymaster 172.16.10.31 1639910615:X 08 Dec 22:52:14.833 # +sdown slave 172.16.10.30:16399 172.16.10.30 16399 @ mymaster 172.16.10.31 16399
172.16.10.30恢复日志
10615:X 08 Dec 22:56:50.802 # -sdown slave 172.16.10.30:16399 172.16.10.30 16399 @ mymaster 172.16.10.31 1639910615:X 08 Dec 22:57:00.751 * +convert-to-slave slave 172.16.10.30:16399 172.16.10.30 16399 @ mymaster 172.16.10.31 16399
6.python client 实现读写分离。
[root@cmdb redis-sentinel]# tree .├── eg.py└── redis_cluster ├── __init__.py ├── __init__.pyc ├── RedisSentinel.py ├── RedisSentinel.pyc ├── settings.py └── settings.pyc1 directory, 7 files[root@cmdb redis-sentinel]#
settings.py是配置文件
[root@cmdb redis-sentinel]# cat redis_cluster/settings.py#default timeoutSOCKET_TIMEOUT = 0.1#sentinel cluster nodes listSENTINEL_NODES = [('192.168.7.30',5001),('192.168.7.31',5001)]#sentinel cluster nameCLUSTER_NAME = "mymaster"#default select db CLUSTER_DB= 15 [root@cmdb redis-sentinel]#
RedisSentinel.py是处理方法
[root@cmdb redis-sentinel]# cat redis_cluster/RedisSentinel.py#!/usr/bin/env python #encoding:utf-8#from redis.sentinel import Sentinelfrom settings import SENTINEL_NODES,SOCKET_TIMEOUT,CLUSTER_NAME,CLUSTER_DBclass RedisSentinelClient: def __init__(self,sentinel_nodes,socket_timeout,cluster_name,db): self.sentinel=Sentinel(sentinel_nodes,socket_timeout=socket_timeout) self.master_name=cluster_name self.socket_timeout=socket_timeout self.db=db def master(self,socket_timeout=None): return self.sentinel.master_for(self.master_name,db=self.db,socket_timeout=self.socket_timeout) def slave(self,socket_timeout=None): return self.sentinel.slave_for(self.master_name,db=self.db,socket_timeout=self.socket_timeout)sentinel_client=RedisSentinelClient(SENTINEL_NODES,SOCKET_TIMEOUT,CLUSTER_NAME,CLUSTER_DB)master_redis=sentinel_client.master()slave_redis=sentinel_client.slave()[root@cmdb redis-sentinel]#
eg.py是使用的方法:
[root@cmdb redis-sentinel]# cat eg.py #!/usr/bin/env python #encoding:utf-8from redis_cluster.RedisSentinel import master_redis,slave_redismaster_redis.set('ceshi','nihaoma') #写入数据print slave_redis.get('ceshi') #读取数据的。此处读取数据是从库上读取的[root@cmdb redis-sentinel]#
到此就实现了python对 sentinel哨兵集群的读写分离操作