Using Multiple Masters
KeyDB now has support with multiple masters when replicating. This mode is enabled with the following configuration line:
When KeyDB connects with multiple masters it behaves differently than with traditional replication:
- Multiple invocations of the replicaof command will result in adding additional masters, not replacing the current one
- KeyDB will not drop its database when sync'ing with the master
- KeyDB will merge any reads/writes from the master with its own internal database
- KeyDB will default to last operation wins
This means that a replica with multiple masters will contain a superset of the data of all its masters. If two masters have a value with the same key it is undefined which key will be taken. If a master deletes a key that exists on another master the replica will no longer contain a copy of that key.
Setting up Multi-Master#
Multiple Master setup is very similar to active replica setup, but allows more than one replica node (all masters). It allows you to read and write to all instances which can increase reads under high load, and have your other master nodes ready to go in a failure scenario of one.
With multi-master setup you make each master a replica of other nodes. This can accept many topologies, you could make different variations of ring topologies or make every master a replica of all other masters. If not all are synced, consider failure scenarios and ensure that one break wont cause others to lose their connections.
An example config file:
Config File#
Note: When setting up multi-master in a config file, make sure to enable multi-master and active-replica BEFORE setting your replicaof commands, as shown in the examples below, otherwise multi-master replication may not work correctly.
Instance-A config file:
Instance-B config file:
Instance-C config file:
You can also append to config file keydb-server --multi-master yes --active-replica yes --replicaof [ipaddress] [port] --replicaof [ipaddress] [port]
Removing A Replica Node#
During operation you may want to remove a node without having to reload configuration files & restart nodes. This can be done using keydb-cli:
Note that this command will need to be issued on all nodes within the cluster (we do not yet support automatic propogation accross cluster nodes).