最近项目上需要对Redis(目前redis用的是2.8版本)做高可用、集群等优化,就扩展了jedis客户端(MasterSlaveJedis、MasterSlaveJedisPool、ShardedMasterSlaveJedis、ShardedMasterSlaveJedisPool),以满足以下几个需求:
1、主Redis与各个从Redis之间能够实现高可用(监控、自动故障迁移等),这个委托给redis-sentinel来实现。
2、业务代码中为分散redis服务的压力,全部写操作走主库,部分读操作需要走从服务器的业务需求。(就目前jedis 2.6.2版本中的JedisSentinelPool做不到这点)
3、对多个Master-Slaves组成的高可用单元做sharding集群
一、针对仅Master-Slaves架构组成的使用例子
sentinel_63791.conf
port 63791
daemonize yes
logfile "/var/log/redis/sentinel_63791.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
sentinel_63792.conf
port 63792
daemonize yes
logfile "/var/log/redis/sentinel_63792.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
二、针对由各个Master-Slaves单元组成的sharding集群架构使用例子
假设有两个Master-Slaves单元:
master-1 : master=192.168.137.101:6379 slaves=[192.168.137.101:6380, 192.168.137.101:6381](相当于一个shard)
master-2 : master=192.168.137.101:6382 slaves=[192.168.137.101:6383, 192.168.137.101:6384](相当于一个shard)
sentinel_63791.conf
port 63791
daemonize yes
logfile "/var/log/redis/sentinel_63791.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
#master-2
sentinel monitor master-2 192.168.137.101 6382 2
sentinel down-after-milliseconds master-2 15000
sentinel parallel-syncs master-2 1
sentinel failover-timeout master-2 90000
sentinel_63792.conf
port 63792
daemonize yes
logfile "/var/log/redis/sentinel_63792.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
#master-2
sentinel monitor master-2 192.168.137.101 6382 2
sentinel down-after-milliseconds master-2 15000
sentinel parallel-syncs master-2 1
sentinel failover-timeout master-2 90000
项目地址:https://github.com/penggle/jedis-ms-sentinel
该项目已经发布到maven中央库上:
maven:
<groupId>com.github.penggle</groupId>
<artifactId>jedis-ms-sentinel</artifactId>
<version>1.0.0</version>
<name>jedis-ms-sentinel</name>
gradle:
compile("com.github.penggle:jedis-ms-sentinel:1.0.0")
已有 0人发表留言,猛击->> 这里<<-参与讨论
ITeye推荐
1、主Redis与各个从Redis之间能够实现高可用(监控、自动故障迁移等),这个委托给redis-sentinel来实现。
2、业务代码中为分散redis服务的压力,全部写操作走主库,部分读操作需要走从服务器的业务需求。(就目前jedis 2.6.2版本中的JedisSentinelPool做不到这点)
3、对多个Master-Slaves组成的高可用单元做sharding集群
一、针对仅Master-Slaves架构组成的使用例子
sentinel_63791.conf
port 63791
daemonize yes
logfile "/var/log/redis/sentinel_63791.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
sentinel_63792.conf
port 63792
daemonize yes
logfile "/var/log/redis/sentinel_63792.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
Set<String> sentinels = new LinkedHashSet<String>(); sentinels.add("192.168.137.101:63791"); sentinels.add("192.168.137.101:63792"); Pool<MasterSlaveJedis> masterSlaveJedisPool = new MasterSlaveJedisSentinelPool("master-1", sentinels,jedisPoolConfig); MasterSlaveJedis masterSlaveJedis = masterSlaveJedisPool.getResource(); //>>> masterSlaveJedis = MasterSlaveJedis {master=192.168.137.101:6379, slaves=[192.168.137.101:6380, 192.168.137.101:6381]} System.out.println(">>> masterSlaveJedis = " + masterSlaveJedis); masterSlaveJedis.set("nowTime", "2015-03-16 15:34:55"); // 底层实际调用主Redis上的相应方法master.set("nowTime", "2015-03-16 15:34:55"); LockSupport.parkNanos(TimeUnit.MILLISECONDS.toNanos(200)); String slaveHolder1 = "myslave1"; Jedis slave1 = masterSlaveJedis.opsForSlave(slaveHolder); //如果全部slave都挂掉的话,那么opsForSlave则返回master System.out.println(">>> nowTime = " + slave1.get("nowTime")); //>>> nowTime = 2015-03-16 15:34:55 String slaveHolder2 = "myslave1"; Jedis slave2 = masterSlaveJedis.opsForSlave(slaveHolder); System.out.println(">>> nowTime = " + slave2.get("nowTime")); //>>> nowTime = 2015-03-16 15:34:55 System.out.println(slave1.equals(slave2)); // 如果slaveHolder1 equals slaveHolder2那么slave1.equals(slave2)肯定是同一个对象 masterSlaveJedisPool.returnResource(masterSlaveJedis);
二、针对由各个Master-Slaves单元组成的sharding集群架构使用例子
假设有两个Master-Slaves单元:
master-1 : master=192.168.137.101:6379 slaves=[192.168.137.101:6380, 192.168.137.101:6381](相当于一个shard)
master-2 : master=192.168.137.101:6382 slaves=[192.168.137.101:6383, 192.168.137.101:6384](相当于一个shard)
sentinel_63791.conf
port 63791
daemonize yes
logfile "/var/log/redis/sentinel_63791.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
#master-2
sentinel monitor master-2 192.168.137.101 6382 2
sentinel down-after-milliseconds master-2 15000
sentinel parallel-syncs master-2 1
sentinel failover-timeout master-2 90000
sentinel_63792.conf
port 63792
daemonize yes
logfile "/var/log/redis/sentinel_63792.log"
#master-1
sentinel monitor master-1 192.168.137.101 6379 2
sentinel down-after-milliseconds master-1 15000
sentinel parallel-syncs master-1 1
sentinel failover-timeout master-1 90000
#master-2
sentinel monitor master-2 192.168.137.101 6382 2
sentinel down-after-milliseconds master-2 15000
sentinel parallel-syncs master-2 1
sentinel failover-timeout master-2 90000
Set<String> masterNames = new LinkedHashSet<String>(); masterNames.add("master-1"); masterNames.add("master-2"); Set<String> sentinels = new LinkedHashSet<String>(); sentinels.add("192.168.137.101:63791"); sentinels.add("192.168.137.101:63792"); Pool<ShardedMasterSlaveJedis> shardedMasterSlaveJedisPool = new ShardedMasterSlaveJedisSentinelPool(masterNames, sentinels,jedisPoolConfig); ShardedMasterSlaveJedis shardedMasterSlaveJedis = shardedMasterSlaveJedisPool.getResource(); for(int i = 0; i < 10; i++){ String key = "shard-" + i; shardedMasterSlaveJedis.set(key, String.valueOf(i)); // shard-0, shard-2, shard-6, shard-8, shard-9各key分布在master-1[192.168.137.101:6379], shard-1, shard-3, shard-4, shard-5, shard-7各key分布在master-2[192.168.137.101:6382] System.out.println(key + " = " + shardedMasterSlaveJedis.get(key)); LockSupport.parkNanos(TimeUnit.MILLISECONDS.toNanos(200)); System.out.println(key + " = " + shardedMasterSlaveJedis.getShard(key).opsForSlave().get(key)); // 从一个Master-Slaves单元中的某一个从Redis实例获取key的值 } shardedMasterSlaveJedisPool.returnResource(shardedMasterSlaveJedis);
项目地址:https://github.com/penggle/jedis-ms-sentinel
该项目已经发布到maven中央库上:
maven:
<groupId>com.github.penggle</groupId>
<artifactId>jedis-ms-sentinel</artifactId>
<version>1.0.0</version>
<name>jedis-ms-sentinel</name>
gradle:
compile("com.github.penggle:jedis-ms-sentinel:1.0.0")
已有 0人发表留言,猛击->> 这里<<-参与讨论
ITeye推荐