3、Redis 集群特性之容错、数据迁移
前言：
该篇中主要讲解一下redis的容错以及数据的迁移(横向拓展)
redis 集群信息
在前面章节中讲到将Node加入到cluster以后打印了如下日志：
[root@localhost src]# ./redis-trib.rb create --replicas 1 192.168.1.103:7000 192.168.1.103:7001 192.168.1.103:7002 192.168.1.103:7003 192.168.1.103:7004 192.168.1.103:7005 >>> Creating cluster Connecting to node 192.168.1.103:7000: OK Connecting to node 192.168.1.103:7001: OK Connecting to node 192.168.1.103:7002: OK Connecting to node 192.168.1.103:7003: OK Connecting to node 192.168.1.103:7004: OK Connecting to node 192.168.1.103:7005: OK >>> Performing hash slots allocation on 6 nodes... Using 3 masters: 192.168.1.103:7000 192.168.1.103:7001 192.168.1.103:7002 Adding replica 192.168.1.103:7003 to 192.168.1.103:7000 Adding replica 192.168.1.103:7004 to 192.168.1.103:7001 Adding replica 192.168.1.103:7005 to 192.168.1.103:7002 M 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 slots:0-5460(5461 slots) master M: f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 slots:5461-10922 (5462 slots) master M: 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 slots:10923-16383 (5461 slots) master S: 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 replicates 4bc092eb4731152d15172b065c74c7a795fe6304 S: 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 replicates f37ec54101536425ce8798e041ad75a582d7e153 S: b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 replicates 7b0ca3978858454051ad572aa816eec450f31a53 Can I set the above configuration? (type 'yes' to accept): yes >>> Nodes configuration updated >>> Assign a different config epoch to each node >>> Sending CLUSTER MEET messages to join the cluster Waiting for the cluster to join... >>> Performing Cluster Check (using node 192.168.1.103:7000) M: 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 slots:0-5460 (5461 slots) master M: f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 slots:5461-10922 (5462 slots) master M: 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 slots:10923-16383 (5461 slots) master M: 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slots: (0 slots) master replicates 4bc092eb4731152d15172b065c74c7a795fe6304 M: 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slots: (0 slots) master replicates f37ec54101536425ce8798e041ad75a582d7e153 M: b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slots: (0 slots) master replicates 7b0ca3978858454051ad572aa816eec450f31a53 [OK] All nodes agree about slots configuration. >>> Check for open slots... >>> Check slots coverage... [OK] All 16384 slots covered.
可以得到如下信息
| 节点名称 |
7000 |
7001 |
7002 |
7003 |
7004 |
7005 |
| NODEID |
4bc092eb4731152d15172b065c74c7a795fe6304 |
f37ec54101536425ce8798e041ad75a582d7e153 |
7b0ca3978858454051ad572aa816eec450f31a53 |
778e649f47fa98f6d1f6b1f1043812c6685dc4a8 |
907feee1b665554cadc64921c7fcb8c05b8a5ab6 |
b2bea8ede402e2112cced7d7cea52127f18edef2 |
| 主从 |
master |
master |
master |
slave |
slave |
slave |
| 所属mater节点 |
|
|
|
7000 |
7002 |
7001 |
| slot |
0-5460 |
5461-10922 |
10923-16383 |
|
|
|
redis集群提供 16384 slot，slot可以理解为存储单元，在一个slot可存放多个key值，集群环境下，16384个slot分配给master节点，这也解释了为什么在slave节点上不能进行写操作。
连接到任意一个节点，查询集群相关信息
192.168.1.103:7001> cluster info cluster_state:ok cluster_slots_assigned:16384 cluster_slots_ok:16384 cluster_slots_pfail:0 cluster_slots_fail:0 cluster_known_nodes:6 cluster_size:3 cluster_current_epoch:6 cluster_my_epoch:2 cluster_stats_messages_sent:10436 cluster_stats_messages_received:10106
查询集群节点信息
192.168.1.103:7001> cluster nodes 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439569459017 4 connected 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439569460528 3 connected 10923-16383 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439569461031 5 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master - 0 1439569460025 1 connected 0-5460 b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439569459520 6 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 myself,master - 0 0 2 connected 5461-10922 192.168.1.103:7001>
该命令能有效观察集群信息！！！
集群容错：选举
增加节点 node-7006
启动该节点
[root@localhost my-redis-cluster]# ls node-7000.conf node-7001.conf node-7002.conf node-7003.conf node-7004.conf node-7005.conf node-7006.conf [root@localhost my-redis-cluster]# redis-server node-7006.conf [root@localhost my-redis-cluster]# ps -ef | grep redis root 2382 1 0 8月14 ? 00:00:05 redis-server *:7000 [cluster] root 2394 1 0 8月14 ? 00:00:05 redis-server *:7001 [cluster] root 2398 1 0 8月14 ? 00:00:05 redis-server *:7002 [cluster] root 2402 1 0 8月14 ? 00:00:05 redis-server *:7003 [cluster] root 2408 1 0 8月14 ? 00:00:05 redis-server *:7004 [cluster] root 2414 1 0 8月14 ? 00:00:05 redis-server *:7005 [cluster] root 2929 1 0 00:29 ? 00:00:00 redis-server *:7006 [cluster] root 2933 2248 0 00:29 pts/0 00:00:00 grep --color=auto redis
此时该节点不在当前的集群中，将该节点加入到集群,使用 cluster meet ip port，注意是在客户端执行该命令哦～～
192.168.1.103:7001> cluster meet 192.168.1.103 7006 OK
重新查询集群信息
192.168.1.103:7001> cluster nodes 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439570083805 4 connected 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439570085318 3 connected 10923-16383 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439570085822 5 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master - 0 1439570085822 1 connected 0-5460 b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439570084814 6 connected c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 master - 0 1439570084309 0 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 myself,master - 0 0 2 connected 5461-10922
接下来看下如何将7006挂到7000下
切换到7006客户端
mac:bin lkl$ ./redis-cli -c -h 192.168.1.103 -p 7006
192.168.1.103:7006> cluster replicate 4bc092eb4731152d15172b065c74c7a795fe6304 OK
再次查询cluster nodes
192.168.1.103:7006> cluster nodes 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439570605532 2 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master - 0 1439570604025 1 connected 0-5460 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439570603524 1 connected b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439570604528 3 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 master - 0 1439570603524 2 connected 5461-10922 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439570604025 3 connected 10923-16383 c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 myself,slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 0 0 connected
redis 集群中提高容错，在master宕机后，会在master下的slave下自主选举， 当选的slave会升级为mater，同时接管master中的数据。
接下来带大家看下redis的选举
在生产环境中，一般都建议将节点数目设置为奇数，enhance容错能力。
7000下现在有两个slave 7003 、7006，这里再增加一个7007挂在7000下，操作类似前面，不再废话，直接把结果展现出来
[root@localhost redis-3.0.3]# cd my-redis-cluster/ [root@localhost my-redis-cluster]# ls node-7000.conf node-7001.conf node-7002.conf node-7003.conf node-7004.conf node-7005.conf node-7006.conf node-7007.conf [root@localhost my-redis-cluster]# redis-server node-7007.conf [root@localhost my-redis-cluster]# ps -ef | grep redis root 2382 1 0 8月14 ? 00:00:07 redis-server *:7000 [cluster] root 2394 1 0 8月14 ? 00:00:07 redis-server *:7001 [cluster] root 2398 1 0 8月14 ? 00:00:07 redis-server *:7002 [cluster] root 2402 1 0 8月14 ? 00:00:07 redis-server *:7003 [cluster] root 2408 1 0 8月14 ? 00:00:07 redis-server *:7004 [cluster] root 2414 1 0 8月14 ? 00:00:07 redis-server *:7005 [cluster] root 2929 1 0 00:29 ? 00:00:01 redis-server *:7006 [cluster] root 3128 1 0 00:47 ? 00:00:00 redis-server *:7007 [cluster] root 3132 2248 0 00:48 pts/0 00:00:00 grep --color=auto redis
192.168.1.103:7006> cluster meet 192.168.1.103 7007 OK 192.168.1.103:7006> cluster nodes 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439571019922 2 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master - 0 1439571020927 1 connected 0-5460 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439571020424 1 connected b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439571020927 3 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 master - 0 1439571019416 2 connected 5461-10922 054b9e2efd44f8704d1aa6851f75f316911b7b4d 192.168.1.103:7007 master - 0 1439571019214 0 connected 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439571020927 3 connected 10923-16383 c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 myself,slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 0 0 connected
mac:bin lkl$ ./redis-cli -c -h 192.168.1.103 -p 7007 192.168.1.103:7007> cluster replicate 4bc092eb4731152d15172b065c74c7a795fe6304 OK 192.168.1.103:7007> cluster nodes b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439571183241 3 connected 054b9e2efd44f8704d1aa6851f75f316911b7b4d 192.168.1.103:7007 myself,slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 0 0 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master - 0 1439571182232 1 connected 0-5460 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439571183241 3 connected 10923-16383 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439571182736 2 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 master - 0 1439571184249 2 connected 5461-10922 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439571182736 1 connected c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 slave 4bc092eb4731152d15172b065c74c7a795fe6304 0 1439571183745 1 connected
经过以上操作，7007将挂在7000下。
模拟7000宕机，直接kill process
[root@localhost my-redis-cluster]# ps -ef | grep redis root 2382 1 0 8月14 ? 00:00:07 redis-server *:7000 [cluster] root 2394 1 0 8月14 ? 00:00:07 redis-server *:7001 [cluster] root 2398 1 0 8月14 ? 00:00:07 redis-server *:7002 [cluster] root 2402 1 0 8月14 ? 00:00:07 redis-server *:7003 [cluster] root 2408 1 0 8月14 ? 00:00:07 redis-server *:7004 [cluster] root 2414 1 0 8月14 ? 00:00:07 redis-server *:7005 [cluster] root 2929 1 0 00:29 ? 00:00:01 redis-server *:7006 [cluster] root 3128 1 0 00:47 ? 00:00:00 redis-server *:7007 [cluster] root 3132 2248 0 00:48 pts/0 00:00:00 grep --color=auto redis [root@localhost my-redis-cluster]# kill -9 2382
mac:bin lkl$ ./redis-cli -c -h 192.168.1.103 -p 7003 192.168.1.103:7003> cluster nodes 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439571326184 3 connected 10923-16383 054b9e2efd44f8704d1aa6851f75f316911b7b4d 192.168.1.103:7007 slave c75edec9024b2ef4397b70fde2d5227aa9135900 0 1439571324671 7 connected c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 master - 0 1439571324671 7 connected 0-5460 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 myself,slave c75edec9024b2ef4397b70fde2d5227aa9135900 0 0 4 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 master - 0 1439571324671 2 connected 5461-10922 b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439571325177 6 connected 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439571326183 5 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master,fail - 1439571226713 1439571225807 1 disconnected
可以看到 slave通过自主选举功能，7006接替了7000的工作，分配了0-5460slot
数据迁移:slot的变迁
随着业务的发展，redis的节点承载的压力也会增大，redis的集群可通过水平横向的拓展，在集群中加入新的master-slave去分担集群中其他节点的压力。由于redis cluster中数据存放在slot中，可以将线上的reids数据
slot迁移到新加入的master-slave。接下来讲述下如何操作slot。
将指定的slot迁移到指定节点id的master上
192.168.1.103:7006> cluster setslot 6 node 7b0ca3978858454051ad572aa816eec450f31a53 OK 192.168.1.103:7006> cluster nodes 907feee1b665554cadc64921c7fcb8c05b8a5ab6 192.168.1.103:7004 slave f37ec54101536425ce8798e041ad75a582d7e153 0 1439574079852 2 connected 4bc092eb4731152d15172b065c74c7a795fe6304 192.168.1.103:7000 master,fail - 1439571226672 1439571224258 1 disconnected 778e649f47fa98f6d1f6b1f1043812c6685dc4a8 192.168.1.103:7003 slave c75edec9024b2ef4397b70fde2d5227aa9135900 0 1439574077836 7 connected b2bea8ede402e2112cced7d7cea52127f18edef2 192.168.1.103:7005 slave 7b0ca3978858454051ad572aa816eec450f31a53 0 1439574078342 3 connected f37ec54101536425ce8798e041ad75a582d7e153 192.168.1.103:7001 master - 0 1439574079852 2 connected 5461-10922 054b9e2efd44f8704d1aa6851f75f316911b7b4d 192.168.1.103:7007 slave c75edec9024b2ef4397b70fde2d5227aa9135900 0 1439574078342 7 connected 7b0ca3978858454051ad572aa816eec450f31a53 192.168.1.103:7002 master - 0 1439574078342 3 connected 6 10923-16383 c75edec9024b2ef4397b70fde2d5227aa9135900 192.168.1.103:7006 myself,master - 0 0 7 connected 0-5 7-5460
这里只是举个例子，迁移还有很多的其他的操作，大家可以自行搜索。