Redis Cluster高可用集群在线迁移操作记录【转】

时间:2022-02-14 20:15:29

之前介绍了redis cluster的结构及高可用集群部署过程,今天这里简单说下redis集群的迁移。由于之前的redis cluster集群环境部署的服务器性能有限,需要迁移到高配置的服务器上。考虑到是线上生产环境,决定在线迁移,迁移过程,不中断服务。操作过程如下:

一、机器环境

1
2
3
4
5
6
7
8
9
10
11
12
13
迁移前机器环境
-------------------------------------------------------------------------------
主机名              ip地址             节点端口
redis-node01       172.16.60.207     7000,7001
redis-node02       172.16.60.208     7002,7003
redis-node03       172.16.60.209     7004,7005
 
迁移后机器环境
-------------------------------------------------------------------------------
主机名             ip地址             节点端口
redis-new01       172.16.60.202     7000,7001
redis-new02       172.16.60.204     7002,7003
redis-new03       172.16.60.205     7004,7005

二、迁移前redis cluster高可用集群环境部署(这里采用"三主三从"模式)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
三台节点机器安装操作如下一致
[root@redis-node01 ~]# yum install -y gcc g++ make gcc-c++ kernel-devel automake autoconf libtool make wget tcl vim ruby rubygems unzip git
[root@redis-node01 ~]# /etc/init.d/iptables stop
[root@redis-node01 ~]# setenforce 0
[root@redis-node01 ~]# vim /etc/sysconfig/selinux
SELINUX=disabled
 
提前做好下面的准备操作,否则redis日志里会有相应报错
[root@redis-node01 ~]# echo "512" > /proc/sys/net/core/somaxconn     
[root@redis-node01 ~]# vim /etc/rc.local
echo "512" > /proc/sys/net/core/somaxconn  
[root@redis-node01 ~]# echo 1 > /proc/sys/vm/overcommit_memory
[root@redis-node01 ~]# sysctl vm.overcommit_memory=1
vm.overcommit_memory = 1
[root@redis-node01 ~]# vim /etc/sysctl.conf
vm.overcommit_memory=1
[root@redis-node01 ~]# echo never > /sys/kernel/mm/transparent_hugepage/enabled
[root@redis-node01 ~]# vim /etc/rc.local
echo never > /sys/kernel/mm/transparent_hugepage/enabled
 
下载并编译安装redis
[root@redis-node01 ~]# mkdir -p /data/software/
[root@redis-node01 software]# wget http://download.redis.io/releases/redis-4.0.6.tar.gz
[root@redis-node01 software]# tar -zvxf redis-4.0.6.tar.gz
[root@redis-node01 software]# mv redis-4.0.6 /data/
[root@redis-node01 software]# cd /data/redis-4.0.6/
[root@redis-node01 redis-4.0.6]# make
     
-------------------------------------------------------------------------------
分别创建和配置节点
节点1配置
[root@redis-node01 ~]# mkdir /data/redis-4.0.6/redis-cluster
[root@redis-node01 ~]# cd /data/redis-4.0.6/redis-cluster
[root@redis-node01 redis-cluster]# mkdir 7000 7001
[root@redis-node01 redis-cluster]# mkdir /var/log/redis
[root@redis-node01 redis-cluster]# vim 7000/redis.conf
port 7000
bind 172.16.60.207
daemonize yes
pidfile /var/run/redis_7000.pid
logfile /var/log/redis/redis_7000.log
cluster-enabled yes
cluster-config-file nodes_7000.conf
cluster-node-timeout 10100
appendonly yes
     
[root@redis-node01 redis-cluster]# vim 7001/redis.conf
port 7001
bind 172.16.60.207
daemonize yes
pidfile /var/run/redis_7001.pid
logfile /var/log/redis/redis_7001.log
cluster-enabled yes
cluster-config-file nodes_7001.conf
cluster-node-timeout 10100
appendonly yes
     
节点2配置
[root@redis-node02 ~]# mkdir /data/redis-4.0.6/redis-cluster
[root@redis-node02 ~]# cd /data/redis-4.0.6/redis-cluster
[root@redis-node02 redis-cluster]# mkdir 7002 7003
[root@redis-node02 redis-cluster]# mkdir /var/log/redis
[root@redis-node02 redis-cluster]# vim 7000/redis.conf
port 7002
bind 172.16.60.208
daemonize yes
pidfile /var/run/redis_7002.pid
logfile /var/log/redis/redis_7002.log
cluster-enabled yes
cluster-config-file nodes_7002.conf
cluster-node-timeout 10100
appendonly yes
     
[root@redis-node02 redis-cluster]# vim 7003/redis.conf
port 7003
bind 172.16.60.208
daemonize yes
pidfile /var/run/redis_7003.pid
logfile /var/log/redis/redis_7003.log
cluster-enabled yes
cluster-config-file nodes_7003.conf
cluster-node-timeout 10100
appendonly yes
     
节点3配置
[root@redis-node03 ~]# mkdir /data/redis-4.0.6/redis-cluster
[root@redis-node03 ~]# cd /data/redis-4.0.6/redis-cluster
[root@redis-node03 redis-cluster]# mkdir 7004 7005
[root@redis-node03 redis-cluster]# mkdir /var/log/redis
[root@redis-node03 redis-cluster]# vim 7004/redis.conf
port 7004
bind 172.16.60.209
daemonize yes
pidfile /var/run/redis_7004.pid
logfile /var/log/redis/redis_7004.log
cluster-enabled yes
cluster-config-file nodes_7004.conf
cluster-node-timeout 10100
appendonly yes
     
[root@redis-node03 redis-cluster]# vim 7005/redis.conf
port 7005
bind 172.16.60.209
daemonize yes
pidfile /var/run/redis_7005.pid
logfile /var/log/redis/redis_7005.log
cluster-enabled yes
cluster-config-file nodes_7005.conf
cluster-node-timeout 10100
appendonly yes
     
-------------------------------------------------------------------------------
分别启动redis服务(这里统一在/data/redis-4.0.6/redis-cluster路径下启动redis服务,即nodes_*.conf等文件也在这个路径下产生)
节点1
[root@redis-node01 redis-cluster]# for((i=0;i<=1;i++)); do /data/redis-4.0.6/src/redis-server /data/redis-4.0.6/redis-cluster/700$i/redis.conf; done
[root@redis-node01 redis-cluster]# ps -ef|grep redis
root      1103     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.207:7000 [cluster]              
root      1105     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.207:7001 [cluster]              
root      1315 32360  0 16:16 pts/1    00:00:00 grep redis
     
节点2
[root@redis-node02 redis-cluster]# for((i=2;i<=3;i++)); do /data/redis-4.0.6/src/redis-server /data/redis-4.0.6/redis-cluster/700$i/redis.conf; done
[root@redis-node02 redis-cluster]# ps -ef|grep redis
root      9446     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.208:7002 [cluster]              
root      9448     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.208:7003 [cluster]              
root      9644  8540  0 16:17 pts/0    00:00:00 grep redis
     
节点3
[root@redis-node01 redis-cluster]# for((i=4;i<=5;i++)); do /data/redis-4.0.6/src/redis-server /data/redis-4.0.6/redis-cluster/700$i/redis.conf; done
[root@redis-node03 ~]# ps -ef|grep redis
root      9486     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.209:7004 [cluster]              
root      9488     1  0 15:19 ?        00:00:03 /data/redis-4.0.6/src/redis-server 172.16.60.209:7005 [cluster]              
root      9686  9555  0 16:17 pts/0    00:00:00 grep redis
     
-------------------------------------------------------------------------------
接着在节点1上安装 Ruby(只需要在其中一个节点上安装即可)
[root@redis-node01 ~]# yum -y install ruby ruby-devel rubygems rpm-build
[root@redis-node01 ~]# gem install redis
     
温馨提示:
在centos6.x下执行上面的"gem install redis"操作可能会报错,坑很多!
默认yum安装的ruby版本是1.8.7,版本太低,需要升级到ruby2.2以上,否则执行上面安装会报错!
     
首先安装rvm(或者直接下载证书:https://pan.baidu.com/s/1slTyJ7n  密钥:7uan   下载并解压后直接执行"curl -L get.rvm.io | bash -s stable"即可)
[root@redis-node01 ~]# curl -L get.rvm.io | bash -s stable          //可能会报错,需要安装提示进行下面一步操作
[root@redis-node01 ~]# curl -sSL https://rvm.io/mpapis.asc | gpg2 --import -      //然后再接着执行:curl -L get.rvm.io | bash -s stable
[root@redis-node01 ~]# find / -name rvm.sh
/etc/profile.d/rvm.sh
[root@redis-node01 ~]# source /etc/profile.d/rvm.sh
[root@redis-node01 ~]# rvm requirements
       
然后升级ruby到2.3
[root@redis-node01 ~]# rvm install ruby 2.3.1
[root@redis-node01 ~]# ruby -v
ruby 2.3.1p112 (2016-04-26 revision 54768) [x86_64-linux]
       
列出所有ruby版本
[root@redis-node01 ~]# rvm list
       
设置默认的版本
[root@redis-node01 ~]# rvm --default use 2.3.1
       
更新下载源
[root@redis-node01 ~]# gem sources --add https://gems.ruby-china.org/ --remove https://rubygems.org
https://gems.ruby-china.org/ added to sources
source https://rubygems.org not present in cache
       
[root@redis-node01 ~]# gem sources
*** CURRENT SOURCES ***
       
https://rubygems.org/
https://gems.ruby-china.org/
       
最后就能顺利安装了
[root@redis-node01 ~]# gem install redis
Successfully installed redis-4.0.6
Parsing documentation for redis-4.0.6
Done installing documentation for redis after 1 seconds
1 gem installed
     
-------------------------------------------------------------------------------
接着创建redis cluster集群(在节点1机器上操作即可)
     
首先手动指定三个master节点。master节点最好分布在三台机器上
[root@redis-node01 ~]# /data/redis-4.0.6/src/redis-trib.rb create  172.16.60.207:7000 172.16.60.208:7002  172.16.60.209:7004
     
然后手动指定上面三个master节点各自的slave节点。slave节点也最好分布在三台机器上
[root@redis-node01 ~]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.208:7003  172.16.60.207:7000
[root@redis-node01 ~]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.209:7005  172.16.60.208:7002
[root@redis-node01 ~]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.207:7001  172.16.60.209:7004
     
然后检查下redis cluster集群状态
[root@redis-node01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.207:7000
>>> Performing Cluster Check (using node 172.16.60.207:7000)
M: 971d05cd7b9bb3634ad024e6aac3dff158c52eee 172.16.60.207:7000
   slots:0-5460 (5461 slots) master
   1 additional replica(s)
S: e7592314869c29375599d781721ad76675645c4c 172.16.60.209:7005
   slots: (0 slots) slave
   replicates 0060012d749167d3f72833d916e53b3445b66c62
S: 52b8d27838244657d9b01a233578f24d287979fe 172.16.60.208:7003
   slots: (0 slots) slave
   replicates 971d05cd7b9bb3634ad024e6aac3dff158c52eee
S: 213bde6296c36b5f31b958c7730ff1629125a204 172.16.60.207:7001
   slots: (0 slots) slave
   replicates e936d5b4c95b6cae57f994e95805aef87ea4a7a5
M: e936d5b4c95b6cae57f994e95805aef87ea4a7a5 172.16.60.209:7004
   slots:10923-16383 (5461 slots) master
   1 additional replica(s)
M: 0060012d749167d3f72833d916e53b3445b66c62 172.16.60.208:7002
   slots:5461-10922 (5462 slots) master
   1 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots。 covered.
   
通过上面可以看出,只有master主节点才占用slots,从节点都是0 slots,也就是说keys数值是在master节点上。
三个master主节点分割了16384 slots。分别是0-5460、5461-10922、10923-16383。
如果有一组master-slave都挂掉,16484 slots不完整,则整个集群服务也就挂了,必须等到这组master-slave节点恢复,则整个集群才能恢复。
如果新加入master主节点,默认是0 slots,需要reshard为新master节点分布数据槽(会询问向移动多少哈希槽到此节点),后面会提到。
  
写入几条测试数据
登录三个master节点上写入数据(登录slave节点上写入数据,发现也会自动跳到master节点上进行写入)
[root@redis-node01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.207 -c -p 7000
172.16.60.207:7000> set test1 test-207
OK
172.16.60.207:7000> set test11 test-207-207
-> Redirected to slot [13313] located at 172.16.60.209:7004
OK
  
[root@redis-node01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.208 -c -p 7002
172.16.60.208:7002> set test2 test-208
OK
172.16.60.208:7002> set test22 test-208-208
-> Redirected to slot [4401] located at 172.16.60.207:7000
OK
  
[root@redis-node01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.209 -c -p 7004
172.16.60.209:7004> set test3 test-209
OK
172.16.60.209:7004> set test33 test-209-209
OK
  
读数据
[root@redis-node01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.207 -c -p 7000
172.16.60.207:7000> get test1
"test-207"
172.16.60.207:7000> get test11
-> Redirected to slot [13313] located at 172.16.60.209:7004
"test-207-207"
172.16.60.209:7004> get test2
-> Redirected to slot [8899] located at 172.16.60.208:7002
"test-208"
172.16.60.208:7002> get test22
-> Redirected to slot [4401] located at 172.16.60.207:7000
"test-208-208"
172.16.60.207:7000> get test3
-> Redirected to slot [13026] located at 172.16.60.209:7004
"test-209"
172.16.60.209:7004> get test33
"test-209-209"
172.16.60.209:7004> 

三、在线迁移

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
三台新机器安装redis步骤省略,和上面一致。
三台新机器的各节点配置和迁移前三台机器一直,只需要修改ip地址即可。路径和端口一致
启动三台新机器的redis节点服务
在新节点redis-new01上安装Ruby,安装过程省略,和上面一直。
 
将三个新节点都添加到之前的集群中。
=====================
先添加主节点
命令格式"redis-trib.rb add-node <新增节点名> < 原集群节点名>"
第一个为新节点IP的master端口,第二个参数为现有的任意节点IP的master端口
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node 172.16.60.202:7000 172.16.60.207:7000
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node 172.16.60.204:7002 172.16.60.207:7000
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node 172.16.60.205:7004 172.16.60.207:7000
 
=====================
再添加新机器的从节点
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.204:7003  172.16.60.202:7000
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.205:7005  172.16.60.204:7002
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb add-node --slave 172.16.60.202:7001  172.16.60.205:7004
 
查看此时集群状态
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.202:7000
 
查看集群的哈希槽slot情况
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb info 172.16.60.202:7000
172.16.60.202:7000 (a0169bec...) -> 0 keys | 0 slots | 1 slaves.
172.16.60.209:7004 (47cde5c7...) -> 3 keys | 5461 slots | 1 slaves.
172.16.60.208:7002 (656fc84a...) -> 1 keys | 5462 slots | 1 slaves.
172.16.60.205:7004 (48cbab90...) -> 0 keys | 0 slots | 1 slaves.
172.16.60.207:7000 (a8fe2d6e...) -> 2 keys | 5461 slots | 1 slaves.
172.16.60.204:7002 (c6a78cfb...) -> 0 keys | 0 slots | 1 slaves.
[OK] 6 keys in 6 masters.
0.00 keys per slot on average.
 
新添加的master节点的slot默认都是为0,master主节点如果没有slots的话,存取数据就都不会被选中!
数据只会存储在master主节点中!
所以需要给新添加的master主节点分配slot,即reshard slot操作。
 
如上根据最后一个新master节点添加成功后显示的slot可知,已有的master节点的slot分配情况为:
172.16.60.207:7000   -->  slots:0-5460 (5461 slots) master
172.16.60.208:7002   -->  slots:5461-10922 (5462 slots) master
172.16.60.209:7004   -->  slots:10923-16383 (5461 slots) master
 
现在开始往新添加的三个master节点分配slot
a)将172.16.60.207:7000的slot全部分配(5461)给172.16.60.202:7000
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.202:7000
........
How many slots do you want to move (from 1 to 16384)? 5461          #分配多少数量的slot。(这里要把172.16.60.207:7000节点的slot都分配完)
What is the receiving node ID? a0169becd97ccca732d905fd762b4d615674f7bd      #上面那些数量的slot被哪个节点接收。这里填写172.16.60.202:7000节点ID
Please enter all the source node IDs.
  Type 'all' to use all the nodes as source nodes for the hash slots.
  Type 'done' once you entered all the source nodes IDs.
Source node #1:971d05cd7b9bb3634ad024e6aac3dff158c52eee          #指从哪个节点分配上面指定数量的slot。这里填写172.16.60.207:7000的ID。如果填写all,则表示从之前所有master节点中抽取上面指定数量的slot。
Source node #2:done                       #填写done
.......
Do you want to proceed with the proposed reshard plan (yes/no)? yes     #填写yes,确认分配
 
==================================================================
可能会遇到点问题,resharding执行中断。然后出现两边都有slot的情况。
Moving slot 4396 from 172.16.60.207:7000 to 172.16.60.202:7000:
Moving slot 4397 from 172.16.60.207:7000 to 172.16.60.202:7000:
Moving slot 4398 from 172.16.60.207:7000 to 172.16.60.202:7000:
Moving slot 4399 from 172.16.60.207:7000 to 172.16.60.202:7000:
Moving slot 4400 from 172.16.60.207:7000 to 172.16.60.202:7000:
Moving slot 4401 from 172.16.60.207:7000 to 172.16.60.202:7000:
[ERR] Calling MIGRATE: ERR Syntax error, try CLIENT (LIST | KILL | GETNAME | SETNAME | PAUSE | REPLY)
 
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.202:7000  
>>> Performing Cluster Check (using node 172.16.60.202:7000)
M: a0169becd97ccca732d905fd762b4d615674f7bd 172.16.60.202:7000
   slots:0-4400 (4401 slots) master
   1 additional replica(s)
.......
M: 971d05cd7b9bb3634ad024e6aac3dff158c52eee 172.16.60.207:7000
   slots:4401-5460 (1060 slots) master
   1 additional replica(s)
 
分析原因:
reshard重新分配slot时报错内容为:Syntax error ,try CLIENT (LIST|KILL|GETNAME|SETNAME|PAUSE|REPLY)
但是迁移没有key-value的槽的时候就会执行成功。 这就说明问题出在了存不存在key-value上!
 
找到reshard的执行过程:发现具体迁移步骤是通过 move_slot函数调用(redis-trib.rb文件中)。
打开move_slot函数,找到具体的迁移代码。
[root@redis-new01 redis-cluster]# cp /data/redis-4.0.6/src/redis-trib.rb /tmp/
[root@redis-new01 redis-cluster]# cat /data/redis-4.0.6/src/redis-trib.rb|grep source.r.client.call
                source.r.client.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:keys,*keys])
                    source.r.client.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys])
 
上面grep出来的source.r.client.call部分则就是redis-trib.rb脚本告知客户端执行迁移带key-value槽的指令。
 
我们会发现该指令的具体调用时,等同于
"client migrate target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys]"
 
问题来了,这条指令在服务器中怎么执行的呢?
它先执行networking.c  文件中的 clientCommand(client *c)
 
根据参数一一比对(if条件语句)。这时候就会发现bug来了!!!clientCommand函数中没有 migrate分支。
所以会返回一个    Syntax error ,try CLIENT (LIST|KILL|GETNAME|SETNAME|PAUSE|REPLY);
这个错误信息告诉你, Client中只有LIST|KILL|GETNAME|SETNAME|PAUSE|REPLY分支。
 
那么怎么去修改实现真正的带key迁移的slot呢?
 
研究源码,cluster.c文件中里面有migrateCommand(client *c)。恍然大悟,故只要将redis-trib.rb文件中迁移语句修改为:
  source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,"replace",:keys,*keys])
  source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys])
 
即不执行clientCommand,直接执行migrateCommand。
 
也就是说,只需要将redis-trib.rb文件中原来的
                source.r.client.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:keys,*keys])
                    source.r.client.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys])
改为
                source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,"replace",:keys,*keys])
                    source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys])
 
问题就解决了!
 
[root@redis-new01 redis-cluster]# cat /data/redis-4.0.6/src/redis-trib.rb |grep  source.r.call
                source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,"replace",:keys,*keys])
                    source.r.call(["migrate",target.info[:host],target.info[:port],"",0,@timeout,:replace,:keys,*keys])
 
这个bug是因为ruby的gem不同造成的,以后5.0版本会抛弃redis-trib.rb。直接使用redis-cli客户端实现集群管理!!
==================================================================
 
redis-trib.rb脚本文件修改后,继续将172.16.60.207:7000剩下的slot全部分配给172.16.60.202:7000
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.202:7000
........
>>> Check for open slots...
[WARNING] Node 172.16.60.202:7000 has slots in importing state (4401).
[WARNING] Node 172.16.60.207:7000 has slots in migrating state (4401).
[WARNING] The following slots are open: 4401
>>> Check slots coverage...
[OK] All 16384 slots covered.
*** Please fix your cluster problems before resharding
 
解决办法:
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.202 -c -p 7000
172.16.60.202:7000> cluster setslot 4401 stable
OK
172.16.60.202:7000>
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.207 -c -p 7000
172.16.60.207:7000> cluster setslot 4401 stable
OK
172.16.60.207:7000>
 
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb fix 172.16.60.202:7000 
.......
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
 
 
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.202:7000
......
How many slots do you want to move (from 1 to 16384)? 1060
What is the receiving node ID? a0169becd97ccca732d905fd762b4d615674f7bd     
Please enter all the source node IDs.
  Type 'all' to use all the nodes as source nodes for the hash slots.
  Type 'done' once you entered all the source nodes IDs.
Source node #1:971d05cd7b9bb3634ad024e6aac3dff158c52eee         
Source node #2:done                      
.......
Do you want to proceed with the proposed reshard plan (yes/no)? yes    
 
然后再check检查集群状态.
发现172.16.60.207:7000节点的5461个slot已经移动到172.16.60.202:7000节点上了。
 [root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.202:7000     
>>> Performing Cluster Check (using node 172.16.60.202:7000)
M: a0169becd97ccca732d905fd762b4d615674f7bd 172.16.60.202:7000
   slots:0-5460 (5461 slots) master
   2 additional replica(s)
........
M: 971d05cd7b9bb3634ad024e6aac3dff158c52eee 172.16.60.207:7000
   slots: (0 slots) master
   0 additional replica(s)
 
b)将172.16.60.208:7002的slot(5462)全部分配给172.16.60.204:7002
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.204:7002
.......
How many slots do you want to move (from 1 to 16384)? 5462
What is the receiving node ID? c6a78cfbb77804c4837963b5f589064b6111457a
Please enter all the source node IDs.
  Type 'all' to use all the nodes as source nodes for the hash slots.
  Type 'done' once you entered all the source nodes IDs.
Source node #1:0060012d749167d3f72833d916e53b3445b66c62
Source node #2:done
.......
Do you want to proceed with the proposed reshard plan (yes/no)? yes
 
c)将172.16.60.209:7004的slot(5461)全部分配给172.16.60.205:7004
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.205:7004
.........
How many slots do you want to move (from 1 to 16384)? 5461
What is the receiving node ID? 48cbab906141dd26241ccdbc38bee406586a8d03
Please enter all the source node IDs.
  Type 'all' to use all the nodes as source nodes for the hash slots.
  Type 'done' once you entered all the source nodes IDs.
Source node #1:e936d5b4c95b6cae57f994e95805aef87ea4a7a5
Source node #2:done
.........
Do you want to proceed with the proposed reshard plan (yes/no)? yes
 
待到三个新节点的master都分配完哈希槽slot之后,再次查看下集群状态
发现迁移之前的那三个master的slot都为0了,slot都对应迁移到新的节点的三个master上了
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.202:7000
>>> Performing Cluster Check (using node 172.16.60.202:7000)
M: a0169becd97ccca732d905fd762b4d615674f7bd 172.16.60.202:7000
   slots:0-5460 (5461 slots) master
   2 additional replica(s)
S: d9671ca6b4235931a2a215cc327a400ad4f9a399 172.16.60.205:7005
   slots: (0 slots) slave
   replicates c6a78cfbb77804c4837963b5f589064b6111457a
M: e936d5b4c95b6cae57f994e95805aef87ea4a7a5 172.16.60.209:7004
   slots: (0 slots) master
   0 additional replica(s)
S: 213bde6296c36b5f31b958c7730ff1629125a204 172.16.60.207:7001
   slots: (0 slots) slave
   replicates 48cbab906141dd26241ccdbc38bee406586a8d03
M: 0060012d749167d3f72833d916e53b3445b66c62 172.16.60.208:7002
   slots: (0 slots) master
   0 additional replica(s)
S: 52b8d27838244657d9b01a233578f24d287979fe 172.16.60.208:7003
   slots: (0 slots) slave
   replicates a0169becd97ccca732d905fd762b4d615674f7bd
M: 48cbab906141dd26241ccdbc38bee406586a8d03 172.16.60.205:7004
   slots:10923-16383 (5461 slots) master
   2 additional replica(s)
S: e7592314869c29375599d781721ad76675645c4c 172.16.60.209:7005
   slots: (0 slots) slave
   replicates c6a78cfbb77804c4837963b5f589064b6111457a
S: 2950f2cb6d960cd48e792f7c82d62d2cd07d20f9 172.16.60.204:7003
   slots: (0 slots) slave
   replicates a0169becd97ccca732d905fd762b4d615674f7bd
M: 971d05cd7b9bb3634ad024e6aac3dff158c52eee 172.16.60.207:7000
   slots: (0 slots) master
   0 additional replica(s)
M: c6a78cfbb77804c4837963b5f589064b6111457a 172.16.60.204:7002
   slots:5461-10922 (5462 slots) master
   2 additional replica(s)
S: 6e663a1bcc3d241ed4d1a9667a0cc92fbe554740 172.16.60.202:7001
   slots: (0 slots) slave
   replicates 48cbab906141dd26241ccdbc38bee406586a8d03
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
 
查看集群slot情况
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb info 172.16.60.202:7000
172.16.60.202:7000 (a0169bec...) -> 2 keys | 5461 slots | 2 slaves.
172.16.60.209:7004 (47cde5c7...) -> 0 keys | 0 slots | 0 slaves.
172.16.60.208:7002 (656fc84a...) -> 0 keys | 0 slots | 0 slaves.
172.16.60.205:7004 (48cbab90...) -> 3 keys | 5461 slots | 2 slaves.
172.16.60.207:7000 (a8fe2d6e...) -> 0 keys | 0 slots | 0 slaves.
172.16.60.204:7002 (c6a78cfb...) -> 1 keys | 5462 slots | 2 slaves.
[OK] 6 keys in 6 masters.
0.00 keys per slot on average.
 
检查下数据,发现测试数据也已经迁移到新的master节点上了
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.202 -c -p 7000
172.16.60.202:7000> get test1
"test-207"
172.16.60.202:7000> get test2
-> Redirected to slot [8899] located at 172.16.60.204:7002
"test-208"
172.16.60.204:7002> get test3
-> Redirected to slot [13026] located at 172.16.60.205:7004
"test-209"
172.16.60.205:7004> get test11
"test-207-207"
172.16.60.205:7004> get test22
-> Redirected to slot [4401] located at 172.16.60.202:7000
"test-208-208"
172.16.60.202:7000> get test33
-> Redirected to slot [12833] located at 172.16.60.205:7004
"test-209-209"
172.16.60.205:7004>

关于reshard重新分配哈希槽slot,除了上面交互式的操作,也可以直接使用如下命令进行操作:

1
# redis-trib.rb reshard --from <node-id> --to <node-id> --slots <number of slots> --yes <host>:<port>

四、迁移完成后,从集群中删除原来的节点

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
a)从集群中删除迁移之前的slave从节点
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.207:7001 213bde6296c36b5f31b958c7730ff1629125a204
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.208:7003 52b8d27838244657d9b01a233578f24d287979fe
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.209:7005 e7592314869c29375599d781721ad76675645c4c
   
b)从集群中删除迁移之前的master主节点。
删除master主节点时需注意下面节点:
-  如果主节点有从节点,需要将从节点转移到其他主节点或提前删除从节点
-  如果主节点有slot,去掉分配的slot,然后再删除主节点。
  
删除master主节点时,必须确保它上面的slot为0,即必须为空!否则可能会导致整个redis cluster集群无法工作!
如果要移除的master节点不是空的,需要先用重新分片命令来把数据移到其他的节点。
另外一个移除master节点的方法是先进行一次手动的失效备援,等它的slave被选举为新的master,并且它被作为一个新的slave被重新加到集群中来之后再移除它。
很明显,如果你是想要减少集群中的master数量,这种做法没什么用。在这种情况下你还是需要用重新分片来移除数据后再移除它。
  
由于已经将原来的三个master主节点的slot全部抽完了,即slot现在都为0,且他们各自的slave节点也已在上面删除
所以这时原来的三个master主节点可以直接从集群中删除
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.207:7000 971d05cd7b9bb3634ad024e6aac3dff158c52eee
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.208:7002 0060012d749167d3f72833d916e53b3445b66c62
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb del-node 172.16.60.209:7004 e936d5b4c95b6cae57f994e95805aef87ea4a7a5
   
最后再次查看下新的redis cluster集群状态
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb check 172.16.60.202:7000                                         
>>> Performing Cluster Check (using node 172.16.60.202:7000)
M: a0169becd97ccca732d905fd762b4d615674f7bd 172.16.60.202:7000
   slots:0-5460 (5461 slots) master
   1 additional replica(s)
S: d9671ca6b4235931a2a215cc327a400ad4f9a399 172.16.60.205:7005
   slots: (0 slots) slave
   replicates c6a78cfbb77804c4837963b5f589064b6111457a
M: 48cbab906141dd26241ccdbc38bee406586a8d03 172.16.60.205:7004
   slots:10923-16383 (5461 slots) master
   1 additional replica(s)
S: 2950f2cb6d960cd48e792f7c82d62d2cd07d20f9 172.16.60.204:7003
   slots: (0 slots) slave
   replicates a0169becd97ccca732d905fd762b4d615674f7bd
M: c6a78cfbb77804c4837963b5f589064b6111457a 172.16.60.204:7002
   slots:5461-10922 (5462 slots) master
   1 additional replica(s)
S: 6e663a1bcc3d241ed4d1a9667a0cc92fbe554740 172.16.60.202:7001
   slots: (0 slots) slave
   replicates 48cbab906141dd26241ccdbc38bee406586a8d03
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
   
   
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb info 172.16.60.202:7000                                        
172.16.60.202:7000 (a0169bec...) -> 2 keys | 5461 slots | 1 slaves.
172.16.60.205:7004 (48cbab90...) -> 3 keys | 5461 slots | 1 slaves.
172.16.60.204:7002 (c6a78cfb...) -> 1 keys | 5462 slots | 1 slaves.
[OK] 6 keys in 3 masters.
0.00 keys per slot on average.
   
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-cli -h 172.16.60.202 -c -p 7000
172.16.60.202:7000> get test1
"test-207"
172.16.60.202:7000> get test11
-> Redirected to slot [13313] located at 172.16.60.205:7004
"test-207-207"
172.16.60.205:7004> get test2
-> Redirected to slot [8899] located at 172.16.60.204:7002
"test-208"
172.16.60.204:7002> get test22
-> Redirected to slot [4401] located at 172.16.60.202:7000
"test-208-208"
172.16.60.202:7000> get test3
-> Redirected to slot [13026] located at 172.16.60.205:7004
"test-209"
172.16.60.205:7004> get test33
"test-209-209"
172.16.60.205:7004>
   
=====================================================
温馨提示:
如果被删除的master主节点的slot不为0,则需要先将被删除master节点的slot抽取完,即取消它的slot分配!
   
假设master主节点172.16.60.207:7000的slot还有2550个,则需要将这2550个slot从172.16.60.207:7000上抽到172.16.60.202:7000上
   
[root@redis-new01 redis-cluster]# /data/redis-4.0.6/src/redis-trib.rb reshard 172.16.60.207:7000
.......
How many slots do you want to move (from 1 to 16384)? 2550               //被删除master的所有slot数量
What is the receiving node ID? a0169becd97ccca732d905fd762b4d615674f7bd       //接收2550的slot的master节点ID,即172.16.60.202:7000的ID
Please enter all the source node IDs.
  Type 'all' to use all the nodes as source nodes for the hash slots.
  Type 'done' once you entered all the source nodes IDs.
Source node #1:971d05cd7b9bb3634ad024e6aac3dff158c52eee        //被删除master节点的ID,即172.16.60.207:7000的ID
Source node #2:done                                                                              //输入done
.......
Do you want to proceed with the proposed reshard plan (yes/no)? yes           //确认操作
   
如上成功取消了master节点的slot分配(即slot为0)之后,它就可以被删除了!
   
温馨提示:
1)新增master节点后,也需要进行reshard操作,不过针对的是新增节点。即"redis-trib.rb reshard 新增节点"。这是slot分配操作!
2)删除master节点前,如果有slot,也需要进行reshard操作,不过针对的是删除节点。即"redis-trib.rb reshard 被删除节点"。这是slot取消操作!

经过测试,应用在redis cluster如上迁移过程中没有受到任何影响!但是要注意,迁移后需要在应用程序里将redis连接地址更新为新的redis地址和端口。

转自

Redis Cluster高可用集群在线迁移操作记录 - 散尽浮华 - 博客园 https://www.cnblogs.com/kevingrace/p/9844310.html

Redis Cluster高可用集群在线迁移操作记录【转】的更多相关文章

  1. Redis Cluster高可用集群在线迁移操作记录

    之前介绍了redis cluster的结构及高可用集群部署过程,今天这里简单说下redis集群的迁移.由于之前的redis cluster集群环境部署的服务器性能有限,需要迁移到高配置的服务器上.考虑 ...

  2. 近千节点的Redis Cluster高可用集群案例&colon;优酷蓝鲸优化实战(摘自高可用架构)

    (原创)2016-07-26 吴建超 高可用架构导读:Redis Cluster 作者建议的最大集群规模 1,000 节点,目前优酷在蓝鲸项目中管理了超过 700 台节点,积累了 Redis Clus ...

  3. Redis创建高可用集群教程【Windows环境】

    模仿的过程中,加入自己的思考和理解,也会有进步和收获. 在这个互联网时代,在高并发和高流量可能随时爆发的情况下,单机版的系统或者单机版的应用已经无法生存,越来越多的应用开始支持集群,支持分布式部署了. ...

  4. kubernetes部署redis主从高可用集群

    1.redis主从高可用集群结构 2.k8s部署有状态的服务选择 对于K8S集群有状态的服务,我们可以选择deployment和statefulset statefulset service& ...

  5. ProxySQL Cluster 高可用集群环境部署记录

    ProxySQL在早期版本若需要做高可用,需要搭建两个实例,进行冗余.但两个ProxySQL实例之间的数据并不能共通,在主实例上配置后,仍需要在备用节点上进行配置,对管理来说非常不方便.但是Proxy ...

  6. MySQL读写分离高可用集群及读操作负载均衡(Centos7)

    目录 概述 keepalived和heartbeat对比 一.环境 二.部署 部署lvs代理和keepalived MySQL+heartbeat+drbd的部署 MySQL主从复制 web服务器及a ...

  7. Redis Cluster 4&period;0高可用集群安装、在线迁移操作记录

    之前介绍了redis cluster的结构及高可用集群部署过程,今天这里简单说下redis集群的迁移.由于之前的redis cluster集群环境部署的服务器性能有限,需要迁移到高配置的服务器上.考虑 ...

  8. Redis高可用集群-哨兵模式(Redis-Sentinel)搭建配置教程【Windows环境】

    No cross,no crown . 不经历风雨,怎么见彩虹. Redis哨兵模式,用现在流行的话可以说就是一个"哨兵机器人",给"哨兵机器人"进行相应的配置 ...

  9. Centos7&period;5基于MySQL5&period;7的 InnoDB Cluster 多节点高可用集群环境部署记录

    一.   MySQL InnoDB Cluster 介绍MySQL的高可用架构无论是社区还是官方,一直在技术上进行探索,这么多年提出了多种解决方案,比如MMM, MHA, NDB Cluster, G ...

随机推荐

  1. Java中的Checked Exception——美丽世界中潜藏的恶魔?

    在使用Java编写应用的时候,我们常常需要通过第三方类库来帮助我们完成所需要的功能.有时候这些类库所提供的很多API都通过throws声明了它们所可能抛出的异常.但是在查看这些API的文档时,我们却没 ...

  2. RSA

    https://en.wikipedia.org/wiki/RSA_(cryptosystem)

  3. oracle 11g 修改默认监听端口1521

    OS: Oracle Linux Server release 5.7 DB: Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - ...

  4. RPi 2B GPIO 测试

    /************************************************************************************** * RPi 2B GPI ...

  5. Node&period;js学习 - Web Server

    Client - 客户端,一般指浏览器,浏览器可以通过 HTTP 协议向服务器请求数据. Server - 服务端,一般指 Web 服务器,可以接收客户端请求,并向客户端发送响应数据. Busines ...

  6. SecureCRT通过Javascrip脚本实现Ctrl&plus;C

    代码如下: # $language = "JScript" # $interface = "1.0" for( i=1 ; i < 3600 ; i++) ...

  7. Cloud9 on Docker镜像发送

    老外有做过几个cloud9的镜像,但是都有各种各样的问题. 小弟在此做了一个docker镜像,特此分享,希望各位大佬多提宝贵意见.谢谢. Cloud9 Online IDE Coding anywhe ...

  8. 开源库Magicodes&period;Storage正式发布

    说明 Magicodes.Storage,是心莱科技团队提供的统一存储库,相关库均使用.NET标准库(netstandard2.0)编写,支持.NET Framework以及.NET Core. 我们 ...

  9. LeetCode 709&period;To Lower Case

    Description Implement function ToLowerCase() that has a string parameter str, and returns the same s ...

  10. BZOJ 1003 - 物流运输 - &lbrack;最短路&plus;dp&rsqb;

    题目链接:https://www.lydsy.com/JudgeOnline/problem.php?id=1003 Time Limit: 10 Sec Memory Limit: 162 MB D ...