Over the past few years the usage of NOSQL databases has grown quite a bit. Part of this popularity is due to the scalability and performance seen with NOSQL solutions, one of those highly performant databases is Redis. Redis is an highly popular open source in memory key - value data store that is currently in use at and highly praised by tech companies such as Twitter, Stack Exchange and Github.
I've been following Redis for quite some time now and finally had a project that had a good use case for it. The setup of Redis was so easy that I thought it would be a good idea to share how to install and configure a Redis instance for Master - Slave replication.
What is Master - Slave Replication?
Master - Slave replication is a data replication concept seen often in the database world. This replication allows for a database service to replicate data written to the master instance to a slave instance. This slave instance could be located within the same facility for scaling out read requests or in another facility in order to mitigate a disaster recovery scenario.
In general slave servers are read only, Redis is no different; however it does have the ability to override this setting however data only written to the slave will be lost during a resynchronization to the master server. Usually a Master server can have many slave servers setup, in the case of Redis a slave server can also serve as the master server for other slaves. This configuration allows one to send all write requests to the single master server and be able to scale out any read requests against many slaves.
Setup the Master Server
Install the Server
In this article we will be installing Redis on Ubuntu 13.10; we can do this using the apt package manager.
ubuntu@redis-master:~$ sudo apt-get install redis-server
Configuration
Before starting Redis we will need to configure a few items within the /etc/redis/redis.conf configuration file.
ubuntu@redis-master:~$ sudo vi /etc/redis/redis.conf
Changing the Listening Interface
By default Redis binds to the localhost IP, this would be fine if we were only using Redis locally. In this example however we are setting up replication and will need to change the bind address to the public IP.
Find:
bind 127.0.0.1
Replace with:
bind <youripaddress>
Example:
bind 10.0.3.61
Adding Password Authentication
Redis does not have a concept of users and permissions that you would see with a traditional database such as MySQL, however Redis does have a simple password based authentication. To require password based authentication we will need to un-comment the requirepass setting.
Find:
# requirepass foobared
Replace with:
requirepass somesuperlongpasswordthathaslotsofspecialcharacters
Restart the Redis Server
At this point after saving the configuration file we will restart the redis-server service.
ubuntu@redis-master:~$ sudo /etc/init.d/redis-server restart
Installation on the Slaves
Setting up a Redis Slave is the same as setting up the master; however we will need to add a few more configuration steps.
Install the Redis Server package
ubuntu@redis-slave:~$ sudo apt-get install redis-server
Configuration
We will again need to modify the Redis server configuration file.
ubuntu@redis-slave:~$ sudo vi /etc/redis/redis.conf
Changing the Listening Interface
Again we will change the bind address to the public IP address of the slave server.
Find:
bind 127.0.0.1
Replace with:
bind <youripaddress>
Adding Password Authentication
While slave systems are read only by default it is a good idea to enable password authentication to prevent clients from reading data without authenticating. This setting is optional, if you do not care about securing the data being read than this can stay commented.
Find:
# requirepass foobared
Replace with:
requirepass somesuperlongpasswordthathaslotsofspecialcharacters
Setting up Replication
The below steps will outline setting up replication from the Slave. Replication only needs to be defined on Slave systems, the Master server does not require any special configuration.
Specifying the Master
In the slave configuration file specify the master server to replicate from. Redis has the ability to replicate from a slave, to set this up you would simply specify the first slaves details in place of the master.
Find:
# slaveof <masterip> <masterport>
Replace With:
slaveof <masterip> <masterport>
Example:
slaveof 10.0.3.61 6379
Authentication
If you set a password earlier for the master server you will need to specify that password via the masterauth setting.
Find:
# masterauth <master-password>
Replace With:
masterauth somesuperlongpasswordthathaslotsofspecialcharacters
Start the Service
At this point we can start the redis-server on the slave.
ubuntu@redis-slave:~$ service redis-server start
Testing Replication
To test the replication we will first login to the master server and set a key-value.
Set the key-value on the Redis Master
To login to the Redis server we will use the redis-cli client. The -h flag specifies the host to connect to, by default the redis-cli will connect to the localhost IP.
ubuntu@redis-master:~$ redis-cli -h 10.0.3.61
Once logged in use the AUTH command with the password specified earlier.
redis 10.0.3.61:6379> AUTH <password>
OK
After authenticating we can now set a key-value pair using the SET command.
redis 10.0.3.61:6379> SET replicated:test true
OK
Get the key-value on the Redis Slave
Now that data has been saved to the master we will login to the Redis Slave server using the same redis-cli command and use the GET command to retrieve the replicated data.
ubuntu@redis-slave:~$ redis-cli -h 10.0.3.251
redis 10.0.3.251:6379> AUTH foobared
OK
Once logged in use the GET command to retrieve the value from the replicated:test key
redis 10.0.3.251:6379> GET replicated:test
"true"
At this point we have installed Redis on two systems and setup replication between them. If you required a third server you could repeat the Slave setup steps and either use the Master or Slave server as the source server.