Project practice, redis cluster technology learning (IX)

3） Add slave node

At the beginning of capacity expansion, we put 6385、6386 Nodes join the cluster , node 6385 Some slots and data are migrated as master nodes , However, compared with other master nodes, there is no slave node at present , Therefore, the node does not have fault switching

Ability to move .

At this time, you need to set the node 6386 As 6385 The slave node , So as to ensure the high availability of the whole cluster . Use cluster replicate{masterNodeId} The command adds the corresponding slave node to the master node , Note that in cluster mode slaveof Add slave node operation is no longer supported . As shown below ：

127.0.0.1:6386>cluster replicate

1a205dd8b2819a00dd1e8b6be40a8e2abe77b756

The slave node initiates full replication to the master node , You also need to update the cluster related of the local node

state , Look at the node 6386 The status confirmation has changed to 6385 The slave node of the node

127.0.0.1:6386>cluster nodes

475528b1bcf8e74d227104a6cf1bf70f00c24aae 127.0.0.1:6386 myself,slave 1a205dd8b2

819a00dd1e8b6be40a8e2abe77b756 0 0 8 connected

1a205dd8b2819a00dd1e8b6be40a8e2abe77b756 127.0.0.1:6385 master - 0

1469779083513 9

connected 0-1365 4096 5462-6826 10923-12287

...

This completes the expansion of the entire cluster , The cluster relationship structure is shown in the figure .

10.4.3 Shrink cluster

Process description ：

1） First, we need to determine whether the offline node has a responsible slot , If it is , Need to move the slot to another section

spot , Ensure the integrity of the whole cluster slot node mapping after the node is offline .

2） When the downline node is no longer responsible for the slot or is itself a slave node , You can notify other nodes in the cluster to forget the offline nodes , When all nodes forget the node, they can shut down normally .

1. Offline migration slot

Offline nodes need to migrate their own slots to other nodes , The principle is the same as that of the previous node expansion migration slot

The process is consistent . For example, let's take 6381 and 6384 Node offline , Node information is as follows ：

127.0.0.1:6381> cluster nodes

40b8d09d44294d2e23c7c768efc8fcd153446746 127.0.0.1:6381 myself,master - 0 0 2

connected

12288-16383

4fa7eac4080f0b667ffeab9b87841da49b84a6e4 127.0.0.1:6384 slave

40b8d09d44294d2e2

3c7c768efc8fcd153446746 0 1469894180780 5 connected

...

6381 It's the master node , Responsible slot （12288-16383）,6384 It's its slave node , Pictured 10-26 Shown . Offline 6381 You need to migrate the responsible slot to other nodes before .

Shrinkage is the opposite of expansion and migration ,6381 Becomes the source node , Other master nodes become target nodes , The source node needs to take responsibility of 4096 The slots migrate evenly to the other master nodes . I'm going to use it directly here redis-trib.rb reshard Command to complete the slot migration . Because of every execution reshard Command can only have one target node , So we need to do 3 Time reshard command , Migrate separately 1365、1365、1366 Slot .