redis集群搭建之二~使用redis-trib.rb方法

关于集群的架构请参考redis集群架构
一、准备阶段
准备阶段略长，若已经熟悉redis cluster，可以跳过
1.需要安装ruby
在开始搭建redis之前，你需要确保你本地有ruby(2.0以上),且该ruby 中有redis 扩展包。
ruby 官网地址
ruby 官网安装教程
一般我都是直接下载安装（即官网教程里的“Building from Source”）安装完毕之后记得确认是否安装成功
//查看本机是否安装ruby,以及ruby版本

//注意这里的2.3.1 以及安装时间2016-05-01
$ ruby -v
ruby 2.3.1p481 (2016-05-01 revision 45883) [universal.x86_64-darwin13]

2.需要有ruby redis包

//验证是否安装好ruby redis扩展包：[这里查看ruby 扩展包有更好的办法请告诉我，我对ruby完全不熟悉]
//如果没有安装好ruby redis,请看"2.1 安装ruby redis 扩展包",
//如果已经安装好，直接看"二 实施阶段"
# find / -name "redis"
/Library/Ruby/Gems/2.3.1/gems/redis-3.0.6/lib/redis

2.1 安装ruby redis 扩展包

//查看你本地是否有gem,如果没有安装，请看"2.2 安装gem",否则请看下面内容
$ gem -v
2.0.14
#gem install redis -v 3.0.6
//因为gems官网的镜像在国外，所以国内网络经常断连。你需要链接一个国内镜像，我用的是https://ruby.taobao.org

//查看现在镜像的来源地址
$ gem sources -l
*** CURRENT SOURCES ***

https://rubygems.org/

/**
 * 增加https://ruby.taobao.org为镜像，并且移除官网的https://rubygems.org/镜像
*/
$ gem sources --add https://ruby.taobao.org/ --remove https://rubygems.org/

//查看现在镜像的来源地址
$ gem sources -l
*** CURRENT SOURCES ***

https://ruby.taobao.org
//安装ruby redis 3.0.6包［这个包需要与你本地的redis包的版本一致哦］
#gem install redis -v 3.0.6
//验证是否安装好redis扩展包：[这里查看ruby 扩展包有更好的办法请告诉我，我对ruby完全不熟悉]
# find / -name "redis"
/Library/Ruby/Gems/2.3.1/gems/redis-3.0.6/lib/redis

2.2 安装gem
rubygems官网下载地址
下载后就按照正常的安装步骤安装即可

//"./configure" 该步骤可以配置安装路径，以及其他参数，请查看帮助
$ ./configure
$ make
$ sudo make install
$ gem -v
2.0.14

3.你需要了解redis的架构
redis架构相关 http://blog.csdn.net/naixiyi/article/details/51335059

4.认识redis.conf文件关于redis集群(redis cluster)的配置
要使redis以集群方式启动，而不是普通单例方式启动，需要更改redis.conf文件以下字段.保存为demo_redis.conf

//这里的port 随意你定，你只要保证该端口7000以及(7000+10000)端口是空闲的即可。前者是用于服务client的端口。后者是用于各个实例相互通信的bus port，默认为服务client端口加10000
port 7000
cluster-enabled yes
cluster-config-file nodes.conf
cluster-node-timeout 5000
appendonly yes

5.认识redis-trib.rb脚本
该脚本的位置在./../redis-3.0.6/src/redis-trib.rb。你可以查看有的各个命令

$ ./../redis-3.0.6/src/redis-trib.rb 
Usage: redis-trib   

  create          host1:port1 ... hostN:portN
                  --replicas 
  check           host:port
  info            host:port
  fix             host:port
                  --timeout 
  reshard         host:port
                  --from 
                  --to 
                  --slots 
                  --yes
                  --timeout 
                  --pipeline 
  rebalance       host:port
                  --weight 
                  --auto-weights
                  --threshold 
                  --use-empty-masters
                  --timeout 
                  --simulate
                  --pipeline 
  add-node        new_host:new_port existing_host:existing_port
                  --slave
                  --master-id 
  del-node        host:port node_id
  set-timeout     host:port milliseconds
  call            host:port command arg arg .. arg
  import          host:port
                  --from 
                  --copy
                  --replace
  help            (show this help)

流程为

二 实施阶段(使用redis-trib.rb方法)

//再次确认
//ruby 已经安装
$ ruby -v
ruby 2.3.1p481 (2016-05-01 revision 45883) [universal.x86_64-darwin13]
//ruby redis 包已经安装
# find / -name "redis"
/Library/Ruby/Gems/2.3.1/gems/redis-3.0.6/lib/redis

//大致路径为
1.创建各个实例的目录
2.启动各个目录内的实例
3.连接各个实例
4.给各个实例分配slot
5.增加实例为master
    5.1启动实例
    5.2增加实例为master
    5.3为新增加的master 再分区（resharding) ,即从其他master移动一些slot
6.增加一个实例为salve
    6.1启动实例
    6.2增加实例为master
7.通过master reshard删除一个master实例
7.1 移除改master 实例的slot
7.2 del-node
8.通过master failover删除一个master实例
8.1 人工备份一个master
8.2 del-node 
9.删除一个salve 实例

1.创建各个实例的目录

//一般以port为目录名
$ ls -al
total 0
drwxr-xr-x  10   admin   340  5  8 16:38 .
drwxrwxr-x+ 81 root           admin  2754  5  1 23:40 ..
drwxr-xr-x   3   admin   102  5  8 16:37 9001
drwxr-xr-x   3   admin   102  5  8 16:38 9002
drwxr-xr-x   3   admin   102  5  8 16:38 9003
drwxr-xr-x   3   admin   102  5  8 16:39 9004
drwxr-xr-x   3   admin   102  5  8 16:39 9005
drwxr-xr-x   3   admin   102  5  8 16:39 9006
drwxr-xr-x@ 25   admin   850  4 26 21:53 redis-3.0.6
//将demo_redis.conf(见上面“4.认识redis.conf文件关于redis集群(redis cluster)的配置”) 复制到各个目录中,例如9001 目录中
$ cp  ./demo_redis.conf ./9001/
$ ls -al  ./9001/
drwxr-xr-x   6   admin    204  5  5 17:45 .
drwxr-xr-x  12   admin    408  5  6 12:23 ..
-rwxr-xr-x   1   admin  41611  5  1 23:50 redis.conf
//将该redis.conf中的port端口号改为9001
//其他各个目录也同样

2.启动各个目录内的实例

/**
 * [一定要再该端口的目录下面哦，再外面目录都是不行的，
 * 即便你将redis-server后面地址也重新写正确，都是会报错的，
 * 原因是redis启动后要在该目录下面生成相关与该端口的 
 * log,aof,rdb，node.*.conf等文件]
 */
$ cd 9001
$ pwd
/Applications/redis-cluster/9001
//redis-server这里的只要是你的redis-server地址就可以，我这里是正好是redis-3.0.6包下面
$ ./../redis-3.0.6/src/redis-server  ./redis.conf
//[验证－是否启动了redis 9001实例]
$ ps -e|grep 9001
15962 ??         0:00.01 ./../redis-3.0.6/src/redis-server *:9001 [cluster] 
15965 ttys005    0:00.00 grep 9001
deMacBook-Pro:9001 $ ./../redis-3.0.6/src/redis-cli -c -p 9001
127.0.0.1:9001> cluster nodes
c33b0b3eea734d962022be568344ba9ec64356a9 :9001 myself,master - 0 0 0 connected

//再示例9002
$ cd ./../9002
$ ./../redis-3.0.6/src/redis-server redis.conf 
$ ps -e|grep 900
15962 ??         0:00.10 ./../redis-3.0.6/src/redis-server *:9001 [cluster] 
15974 ??         0:00.01 ./../redis-3.0.6/src/redis-server *:9002 [cluster] 
15976 ttys005    0:00.00 grep 900         
$ ./../redis-3.0.6/src/redis-cli -c -p 9002
127.0.0.1:9002> cluster nodes
296cb822111108b5d67a4753339a936aee1be59a :9002 myself,master - 0 0 0 connected
//其他目录也同理

3.连接各个实例

$ ps -e|grep 900
15962 ??         0:00.68 ./../redis-3.0.6/src/redis-server *:9001 [cluster] 
15974 ??         0:00.62 ./../redis-3.0.6/src/redis-server *:9002 [cluster] 
15989 ??         0:00.42 ./../redis-3.0.6/src/redis-server *:9003 [cluster] 
15997 ??         0:00.40 ./../redis-3.0.6/src/redis-server *:9004 [cluster] 
16030 ??         0:00.06 ./../redis-3.0.6/src/redis-server *:9005 [cluster] 
16039 ??         0:00.01 ./../redis-3.0.6/src/redis-server *:9006 [cluster] 
16034 ttys004    0:00.00 src/redis-cli -c -p 9005
16041 ttys005    0:00.00 grep 900
/*连接*/
$ ./../redis-3.0.6/src/redis-trib.rb create --replicas 1 127.0.0.1:9001 127.0.0.1:9002 127.0.0.1:9003  127.0.0.1:9004  127.0.0.1:9005  127.0.0.1:9006
>>> Creating cluster
>>> Performing hash slots allocation on 6 nodes...
Using 3 masters:
127.0.0.1:9001
127.0.0.1:9002
127.0.0.1:9003
Adding replica 127.0.0.1:9004 to 127.0.0.1:9001
Adding replica 127.0.0.1:9005 to 127.0.0.1:9002
Adding replica 127.0.0.1:9006 to 127.0.0.1:9003
......
   replicates 7bda0805a4172de778f1a7d30a5c9d851c563812
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.

//[验证] 集群是正常可写的
127.0.0.1:9001> set test test
-> Redirected to slot [6918] located at 127.0.0.1:9002
OK

4.给各个实例分配slot

/**
 1. 在client 查看实例状态,可以看到有3个master,3个slave
 2. master还有各自的slot分布情况,例如9003 slots:10923-16383 说明已经分配好slots，就不需要分配了
 */
$ src/redis-cli -c -p 9001
127.0.0.1:9001> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462698245682 5 connected
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462698244170 4 connected
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 master - 0 1462698247198 2 connected 5461-10922
7bda0805a4172de778f1a7d30a5c9d851c563812 127.0.0.1:9003 master - 0 1462698246185 3 connected 10923-16383
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 myself,master - 0 0 1 connected 0-5460
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 slave 7bda0805a4172de778f1a7d30a5c9d851c563812 0 1462698245178 6 connected

//测试9001实例,从上面可以看到9001管理的slots为0-5460
//a9001是在卡槽2641上的，即是在9001node 管理的卡槽
127.0.0.1:9001> cluster keyslot a9001
(integer) 2641
127.0.0.1:9001> set a9001 9001
OK
127.0.0.1:9001> keys *
1) "a9001"
127.0.0.1:9001> cluster keyslot test
(integer) 6918
//测试9002实例，从上面看到9002 管理的slots为5461-10922
127.0.0.1:9001> cluster keyslot test2
(integer) 8899
//显然命中9002管理的卡槽slot 8899  然后客户端也跳转到了9002
127.0.0.1:9001> set test2 9002
-> Redirected to slot [8899] located at 127.0.0.1:9002
OK
127.0.0.1:9002> keys *
1) "test"
2) "test2"
//同理测试9003

5.增加实例为master
5.1 启动实例

//一定要在该实例的目录下启动
$ cd ../9007
$ pwd
/Applications/redis-cluster/9007
//启动实例
$ ./../redis-3.0.6/src/redis-server redis.conf
//*[验证实例已经启动]*/
$ ps -e|grep 9007
16071 ??         0:00.01 ./../redis-3.0.6/src/redis-server *:9007 [cluster] 
$ ./../redis-3.0.6/src/redis-cli -c -p 9007
127.0.0.1:9007> cluster nodes
74b02117a30268842e0ce617b235ee48e3152d04 :9007 myself,master - 0 0 0 connected

5.2增加实例为master

$ ./../redis-3.0.6/src/redis-trib.rb  add-node 127.0.0.1:9007 127.0.0.1:9001
>>> Adding node 127.0.0.1:9007 to cluster 127.0.0.1:9001
>>> Performing Cluster Check (using node 127.0.0.1:9001)
M: c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001
......
[OK] New node added correctly.

5.3为新增加的master 再分区（resharding) ,即从其他master移动一些slot

//你可以在client 查看现在9007的实例是没有slots的
127.0.0.1:9007> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462699050020 2 connected
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 slave 7bda0805a4172de778f1a7d30a5c9d851c563812 0 1462699049010 3 connected
74b02117a30268842e0ce617b235ee48e3152d04 127.0.0.1:9007 myself,master - 0 0 0 connected
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 master - 0 1462699045986 1 connected 0-5460
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462699046994 1 connected
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 master - 0 1462699048003 2 connected 5461-10922
7bda0805a4172de778f1a7d30a5c9d851c563812 127.0.0.1:9003 master - 0 1462699044980 3 connected 10923-16383
//我们从9001 移动200个slots 给9007
$ ./../redis-3.0.6/src/redis-trib.rb reshard --from c33b0b3eea734d962022be568344ba9ec64356a9 \
--to 74b02117a30268842e0ce617b235ee48e3152d04 \
--slots 200 --yes --timeout 5000 127.0.0.1:9001 
>>> Performing Cluster Check (using node 127.0.0.1:9001)
......
Moving slot 199 from 127.0.0.1:9001 to 127.0.0.1:9007: 

6.增加一个实例为salve
6.1启动实例

$ cd ../9008
$ pwd
/Applications/redis-cluster/9008
$ ./../redis-3.0.6/src/redis-server redis.conf
//
//[验证]
$ ps -e|grep 9008
16092 ??         0:00.01 ./../redis-3.0.6/src/redis-server *:9008 [cluster] 
16094 ttys005    0:00.00 grep 9008
$ ./../redis-3.0.6/src/redis-cli -c -p 9008
127.0.0.1:9008> cluster nodes
b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d :9008 myself,master - 0 0 0 connected

6.2增加实例为slave

//增加9008实例为9001(即该c33b...6a9node点)实例的slave
$ ./../redis-3.0.6/src/redis-trib.rb  add-node \
--slave \
--master-id c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9008 \
127.0.0.1:9001
>>> Adding node 127.0.0.1:9008 to cluster 127.0.0.1:9001
...
[OK] New node added correctly.

7.通过master reshard删除一个master实例
我们试着删除9007节点
7.1 移除该master 实例的slot

//这里"74b02..."是9007 node，而"c33b0b..."是9001 node
$ ./../redis-3.0.6/src/redis-trib.rb reshard \
--from 74b02117a30268842e0ce617b235ee48e3152d04 \
--to c33b0b3eea734d962022be568344ba9ec64356a9 \
--slots 200  \
--yes \
127.0.0.1:9001
>>> Performing Cluster Check (using node 127.0.0.1:9001)
......
Moving slot 199 from 127.0.0.1:9007 to 127.0.0.1:9001: 

7.2 del-node

//查看cluster nodes 发现 9007已经没有slots
127.0.0.1:9002> cluster nodes
74b02117a30268842e0ce617b235ee48e3152d04 127.0.0.1:9007 master - 0 1462700615149 7 connected
//删除9007 无slots的master 节点
$ ./../redis-3.0.6/src/redis-trib.rb del-node 127.0.0.1:9001 74b02117a30268842e0ce617b235ee48e3152d04
>>> Removing node 74b02117a30268842e0ce617b235ee48e3152d04 from cluster 127.0.0.1:9001
>>> Sending CLUSTER FORGET messages to the cluster...
>>> SHUTDOWN the node.

//[验证]
$ ps -e|grep 9007
16147 ttys005    0:00.00 grep 9007
//查看nodes表 已经无9007
127.0.0.1:9002> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462700880271 5 connected
b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d 127.0.0.1:9008 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462700881277 8 connected
7bda0805a4172de778f1a7d30a5c9d851c563812 127.0.0.1:9003 master - 0 1462700879261 3 connected 10923-16383
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462700876236 8 connected
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 slave 7bda0805a4172de778f1a7d30a5c9d851c563812 0 1462700876742 6 connected
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 myself,master - 0 0 2 connected 5461-10922
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 master - 0 1462700877749 8 connected 0-5460

8.通过master failover删除一个master实例
我们这里删除9003，使9003 的从库slave 9006 failover 它。
8.1 人工备份一个master

//使9006 备份9003，成为新的master
/**
 * This command, that can only be send to a Redis Cluster slave node, forces the slave to start a manual 
 * failover of its master instance.
 * 这个命令必须在slave node上面执行，使slave去failover它的master(小三去赶下去正室，从而成为正室)
 * */
 $ ./../redis-3.0.6/src/redis-cli -c -p 9006
127.0.0.1:9006> cluster failover
OK
//[验证]
127.0.0.1:9006> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462701334986 5 connected
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462701332974 8 connected
b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d 127.0.0.1:9008 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462701335994 8 connected
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 master - 0 1462701331964 8 connected 0-5460
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 myself,master - 0 0 9 connected 10923-16383
7bda0805a4172de778f1a7d30a5c9d851c563812 127.0.0.1:9003 slave aa74ab2bb7a31f74a3b9b30efb4090d39472742d 0 1462701337004 9 connected
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 master - 0 1462701333978 2 connected 5461-10922

8.2 del-node

//
//删除9003
$ ./../redis-3.0.6/src/redis-trib.rb del-node 127.0.0.1:9001 7bda0805a4172de778f1a7d30a5c9d851c563812
>>> Removing node 7bda0805a4172de778f1a7d30a5c9d851c563812 from cluster 127.0.0.1:9001
>>> Sending CLUSTER FORGET messages to the cluster...
>>> SHUTDOWN the node.

//[验证] 已经删除9003 node
127.0.0.1:9006> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462701613370 5 connected
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462701609338 8 connected
b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d 127.0.0.1:9008 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462701610349 8 connected
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 master - 0 1462701614375 8 connected 0-5460
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 myself,master - 0 0 9 connected 10923-16383
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 master - 0 1462701612363 2 connected 5461-10922

9.删除一个salve 实例

//
//删除9008 slave node,node-id:7bda0805a...
$ ./../redis-3.0.6/src/redis-trib.rb del-node 127.0.0.1:9001 7bda0805a4172de778f1a7d30a5c9d851c563812
>>> Removing node 7bda0805a4172de778f1a7d30a5c9d851c563812 from cluster 127.0.0.1:9001
>>> Sending CLUSTER FORGET messages to the cluster...
>>> SHUTDOWN the node.
zhengshanshandeMacBook-Pro:9008 zhengshanshan$ ./../redis-3.0.6/src/redis-trib.rb del-node 127.0.0.1:9001 b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d
>>> Removing node b88cea86d04a4a3d2fb60d6c97284b5bd06e6a5d from cluster 127.0.0.1:9001
>>> Sending CLUSTER FORGET messages to the cluster...
>>> SHUTDOWN the node.

//[验证] 已经删除9008 node
$ ps -e|grep 9008
16183 ttys005    0:00.00 grep 9008
127.0.0.1:9006> cluster nodes
b1953bf7307e0d844ecf4cc34256aea791e00370 127.0.0.1:9005 slave 296cb822111108b5d67a4753339a936aee1be59a 0 1462701683953 5 connected
16e79f7ceb0bfd6fb128200ad8c12b4f72a36176 127.0.0.1:9004 slave c33b0b3eea734d962022be568344ba9ec64356a9 0 1462701686980 8 connected
c33b0b3eea734d962022be568344ba9ec64356a9 127.0.0.1:9001 master - 0 1462701684962 8 connected 0-5460
aa74ab2bb7a31f74a3b9b30efb4090d39472742d 127.0.0.1:9006 myself,master - 0 0 9 connected 10923-16383
296cb822111108b5d67a4753339a936aee1be59a 127.0.0.1:9002 master - 0 1462701685972 2 connected 5461-10922