4.1 Exercise - Creating a Redis Cluster

To create a cluster, we need to spin up a few empty Redis instances and configure them to run in cluster mode.

Here’s a minimal configuration file for Redis Cluster:

On the first line we specify the port on which the server should run, then we state that we want the server to run in cluster mode, with the cluster-enabled yes directive. cluster-config-file defines the name of the file where the configuration for this node is stored, in case of a server restart. Finally, cluster-node-timeout is the number of milliseconds a node must be unreachable for it to be considered in failure state.

Let’s create a cluster on your localhost with three primary shards and three replicas (remember, in production always use two replicas to protect against a split-brain situation). We’ll need to bring up six Redis processes and create a redis.conf file for each of them, specifying their port and the rest of the configuration directives above.

First, create six directories:

Then create the minimal configuration redis.conf file from above in each one of them, making sure you change the port directive to match the directory name.

To copy the initial redis.conf file to each folder, run the following:

You should end up with the following directory structure:

Open six terminal tabs and start the servers by going into each one of the directories and starting a Redis instance:

Now that you have six empty Redis servers running, you can join them in a cluster:

Here we list the ports and IP addresses of all six servers and use the create command to instruct Redis to join them in a cluster, creating one replica for each primary. redis-cli will propose a configuration; accept it by typing yes. The cluster will be configured and joined, which means, instances will be bootstrapped into talking with each other.

Finally, you should see a message saying:

This means that there is at least a master instance serving each of the 16384 slots available.

Let’s add a new shard to the cluster, which is something you might do when you need to scale.

First, as before, we need to start two new empty Redis instances (primary and its replica) in cluster mode. We create new directories 7006 and 7007 and in them we copy the same redis.conf file we used before, making sure we change the port directive in them to the appropriate port (7006 and 7007).

Update the port numbers in the files ./7006/redis.conf and ./7007/redis.conf to 7006 and 7007, respectively.

Let’s start the Redis instances:

In the next step we join the new primary shard to the cluster with the add-node command. The first parameter is the address of the new shard, and the second parameter is the address of any of the current shards in the cluster.

Finally we need to join the new replica shard, with the same add-node command, and a few extra arguments indicating the shard is joining as a replica and what will be its primary shard. If we don’t specify a primary shard Redis will assign one itself.

We can find the IDs of our shards by running the cluster nodes command on any of the shards:

The port of the primary shard we added in the last step was 7006, and we can see it on the first line. It’s id is 46a768cfeadb9d2aee91ddd882433a1798f53271.

The resulting command is:

The flag cluster-slave indicates that the shard should join as a replica and --cluster-master-id 46a768cfeadb9d2aee91ddd882433a1798f53271 specifies which primary shard it should replicate.

Now our cluster has eight shards (four primary and four replica), but if we run the cluster slots command we’ll see that the newly added shards don’t host any hash slots, and thus - data. Let’s assign some hash slots to them:

We use the command reshard and the address of any shard in the cluster as an argument here. In the next step we’ll be able to choose the shards we’ll be moving slots from and to.

The first question you’ll get is about the number of slots you want to move. If we have 16384 slots in total, and four primary shards, let’s get a quarter of all shards, so the data is distributed equally. 16384 ÷ 4 is 4096, so let’s use that number.

The next question is about the receiving shard id; the ID of the primary shard we want to move the data to, which we learned how to get in the previous Step, with the cluster nodes command.

Finally, we need to enter the IDs of the shards we want to copy data from. Alternatively, we can type “all” and the shard will move a number of hash slots from all available primary shards.

Once the command finishes we can run the cluster slots command again and we’ll see that our new primary and replica shards have been assigned some hash slots: