redis cluster

redis cluster:

redis3.0版本出现了redis集群功能

架构细节:

(1)所有的redis节点彼此互联(PING-PONG机制),内部使用二进制协议优化传输速度和带宽.

(2)节点的fail是通过集群中超过半数的节点检测失效时才生效.

(3)客户端与redis节点直连,不需要中间proxy层.客户端不需要连接集群所有节点,连接集群中任何一个可用节点即可

(4)redis-cluster把所有的物理节点映射到[0-16383]slot上,cluster 负责维护node<->slot<->value

(5)领着选举过程是集群中所有master参与,如果半数以上master节点与master节点通信超过(cluster-node-timeout),认为当前master节点挂掉.

(6):什么时候整个集群不可用(cluster_state:fail),当集群不可用时,所有对集群的操作做都不可用，收到((error) CLUSTERDOWN The cluster is down)错误

    a:如果集群任意master挂掉,且当前master没有slave.集群进入fail状态,也可以理解成进群的slot映射[0-16383]不完成时进入fail状态.

    b:如果进群超过半数以上master挂掉，无论是否有slave集群进入fail状态.

针对事务多个keys映射到同一个node的情况，redis采用hash tags处理方式，主要是针对key的构建形式做处理，规则如下

• IF the key contains a { character.
• AND IF there is a } character to the right of {
• AND IF there are one or more characters between the first occurrence of { and the first occurrence of }.

例子:

• The two keys {user1000}.following and {user1000}.followers will hash to the same hash slot since only the substring user1000 will be hashed in order to compute the hash slot.
• For the key foo{}{bar} the whole key will be hashed as usually since the first occurrence of { is followed by } on the right without characters in the middle.
• For the key foo{{bar}}zap the substring {bar will be hashed, because it is the substring between the first occurrence of { and the first occurrence of } on its right.
• For the key foo{bar}{zap} the substring bar will be hashed, since the algorithm stops at the first valid or invalid (without bytes inside) match of { and }.
• What follows from the algorithm is that if the key starts with {}, it is guaranteed to be hashed as a whole. This is useful when using binary data as key names.

具体的搭建过程可以参考redis官方文档。