Redis Cluster在线迁移

由于之前的redis cluster物理硬件性能不足。决定升级到更好的服务器上。
考虑到redis是核心生产数据库，决定在线迁移，迁移过程，不中断服务。

下面是测试环境的完成迁移步骤：
1. 原环境（测试环境，没有创建slave）

10.21.14.251:7000
10.21.14.251:7001
10.21.14.251:7002

 

2. 在新主机上，启动三个redis实例

10.21.10.120:7000
10.21.10.120:7001
10.21.10.120:7002

 

3. 将三个redis，都添加到集群中. 命令格式redis-trib.rb add-node <新增节点名> < 原集群节点名>

./redis-trib.rb add-node  10.21.10.120:7000  10.21.14.251:7000
./redis-trib.rb add-node  10.21.10.120:7001  10.21.14.251:7000
./redis-trib.rb add-node  10.21.10.120:7002  10.21.14.251:7000

 

4. 确认添加成功之后，开始reshard slot

./redis-trib.rb reshard 10.21.10.120:7000

 

5. 遇到点问题，由于网络超时等原因，导致resharding中断。然后出现两边都有slot的情况，需要通过fix的方式来修复

[redis@ip-10-21-14-251 redis]$ ./redis-trib.rb reshard 10.21.10.120:7000
>>> Performing Cluster Check (using node 10.21.10.120:7000)
M: 4422ab38377fa8828e0f7884570b3b482a66496b 10.21.10.120:7000
   slots:5026-5460 (435 slots) master
   0 additional replica(s)
M: 5b38e63a1091baa3a871a52275489a2aa1d28bfb 10.21.10.120:7002
   slots:894-3397 (2504 slots) master
   0 additional replica(s)
M: bb1572074d41254e5b4d5aae5c52e54f5129d6d5 10.21.14.251:7001
   slots:3398-4999,5461-15922 (12064 slots) master
   0 additional replica(s)
M: 396a7fbd2ec61752f9e848a1d8cc7b405aef0356 10.21.14.251:7000
   slots: (0 slots) master
   0 additional replica(s)
M: 9f215e7e4b511f3d2bbf5d734731899b71a62a3b 10.21.10.120:7001
   slots:0-893,5000-5025,15923-15948 (946 slots) master
   0 additional replica(s)
M: 0c9b383f65ae4fefc5e02617fb76a845d7510a53 10.21.14.251:7002
   slots:15949-16383 (435 slots) master
   0 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
[WARNING] Node 10.21.10.120:7002 has slots in importing state (3398).
[WARNING] Node 10.21.14.251:7001 has slots in migrating state (3398).
[WARNING] The following slots are open: 3398
>>> Check slots coverage...
[OK] All 16384 slots covered.
*** Please fix your cluster problems before resharding     <<<<<<<<<<<<<<<<<<<<<

 

6. 可以使用下面命令进行集群检查

./redis-trib.rb check 10.21.10.120:7000

 

7. 执行下面命令进行修复，然后就可以继续reshard slot了

./redis-trib.rb fix 10.21.10.120:7000

 

8. 全部slot迁移完成

$ ./redis-trib.rb check 10.21.10.120:7000
>>> Performing Cluster Check (using node 10.21.10.120:7000)
M: 4422ab38377fa8828e0f7884570b3b482a66496b 10.21.10.120:7000
   slots:3399-4999,5026-7332 (3908 slots) master
   0 additional replica(s)
M: 5b38e63a1091baa3a871a52275489a2aa1d28bfb 10.21.10.120:7002
   slots:894-3398,7333-15332,15949-16383 (10940 slots) master
   0 additional replica(s)
M: bb1572074d41254e5b4d5aae5c52e54f5129d6d5 10.21.14.251:7001
   slots: (0 slots) master
   0 additional replica(s)
M: 396a7fbd2ec61752f9e848a1d8cc7b405aef0356 10.21.14.251:7000
   slots: (0 slots) master
   0 additional replica(s)
M: 9f215e7e4b511f3d2bbf5d734731899b71a62a3b 10.21.10.120:7001
   slots:0-893,5000-5025,15333-15948 (1536 slots) master
   0 additional replica(s)
M: 0c9b383f65ae4fefc5e02617fb76a845d7510a53 10.21.14.251:7002
   slots: (0 slots) master
   0 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.

 

确认一下状态

[redis@ip-10-21-14-251 redis]$ ./redis-cli -p 7000 cluster nodes
bb1572074d41254e5b4d5aae5c52e54f5129d6d5 10.21.14.251:7001 master - 0 1509611814919 6 connected
9f215e7e4b511f3d2bbf5d734731899b71a62a3b 10.21.10.120:7001 master - 0 1509611811917 14 connected 0-893 5000-5025 15333-15948
5b38e63a1091baa3a871a52275489a2aa1d28bfb 10.21.10.120:7002 master - 0 1509611815923 13 connected 894-3398 7333-15332 15949-16383
396a7fbd2ec61752f9e848a1d8cc7b405aef0356 10.21.14.251:7000 myself,master - 0 0 1 connected
4422ab38377fa8828e0f7884570b3b482a66496b 10.21.10.120:7000 master - 0 1509611813919 12 connected 3399-4999 5026-7332
0c9b383f65ae4fefc5e02617fb76a845d7510a53 10.21.14.251:7002 master - 0 1509611812917 3 connected

 

9. 确认状态OK的话，开始删除节点

 ./redis-trib.rb del-node 10.21.14.251:7000 396a7fbd2ec61752f9e848a1d8cc7b405aef0356
 ./redis-trib.rb del-node 10.21.14.251:7001 bb1572074d41254e5b4d5aae5c52e54f5129d6d5
 ./redis-trib.rb del-node 10.21.14.251:7002 0c9b383f65ae4fefc5e02617fb76a845d7510a53

 

10. 干掉10.21.14.251:7002成功,

 [redis@ip-10-21-14-251 redis]$ ./redis-trib.rb del-node 10.21.14.251:7002 0c9b383f65ae4fefc5e02617fb76a845d7510a53
>>> Removing node 0c9b383f65ae4fefc5e02617fb76a845d7510a53 from cluster 10.21.14.251:7002
>>> Sending CLUSTER FORGET messages to the cluster...
>>> SHUTDOWN the node.
[envuser@ip-10-21-14-251 redis]$ ./redis-trib.rb check 10.21.10.120:7000
>>> Performing Cluster Check (using node 10.21.10.120:7000)
M: 4422ab38377fa8828e0f7884570b3b482a66496b 10.21.10.120:7000
   slots:3399-4999,5026-7332 (3908 slots) master
   0 additional replica(s)
M: 5b38e63a1091baa3a871a52275489a2aa1d28bfb 10.21.10.120:7002
   slots:894-3398,7333-15332,15949-16383 (10940 slots) master
   0 additional replica(s)
M: bb1572074d41254e5b4d5aae5c52e54f5129d6d5 10.21.14.251:7001
   slots: (0 slots) master
   0 additional replica(s)
M: 396a7fbd2ec61752f9e848a1d8cc7b405aef0356 10.21.14.251:7000
   slots: (0 slots) master
   0 additional replica(s)
M: 9f215e7e4b511f3d2bbf5d734731899b71a62a3b 10.21.10.120:7001
   slots:0-893,5000-5025,15333-15948 (1536 slots) master
   0 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.

 

状态检查

[redis@ip-10-21-14-251 redis]$ ./redis-cli -p 7000 cluster nodes
bb1572074d41254e5b4d5aae5c52e54f5129d6d5 10.21.14.251:7001 master - 0 1509611900094 6 connected
9f215e7e4b511f3d2bbf5d734731899b71a62a3b 10.21.10.120:7001 master - 0 1509611902100 14 connected 0-893 5000-5025 15333-15948
5b38e63a1091baa3a871a52275489a2aa1d28bfb 10.21.10.120:7002 master - 0 1509611901098 13 connected 894-3398 7333-15332 15949-16383
396a7fbd2ec61752f9e848a1d8cc7b405aef0356 10.21.14.251:7000 myself,master - 0 0 1 connected
4422ab38377fa8828e0f7884570b3b482a66496b 10.21.10.120:7000 master - 0 1509611899093 12 connected 3399-4999 5026-7332

 

根据上面步骤，删除剩余节点即可。
经过测试，应用在迁移过程中，没有受到任何影响。但是应用连接池的IP需要找机会增加10.21.10.120。