This repo contains a minimal setup for a Redis cluster using Docker Compose. For a kubernetes setup, I recommend using this repo as a guide.

Setup

git clone https://github.com/jhsul/redis-cluster-docker-compose
cd redis-cluster-docker-compose

# Start the cluster
docker compose up -d

# (First time only) Configure the cluster
docker compose up redis-cluster-init

The redis nodes will remember their cluster configuration in their container filesystem even if they are restarted. If you want to start from scratch, you can remove all the containers:

docker compose stop redis-cluster
docker compose rm redis-cluster

Usage

You can access the cluster using redis-cli directly from any of the cluster nodes.

docker exec redis-cluster-docker-compose-redis-cluster-1 redis-cli -c cluster nodes

# Output
692224203532603c8b1103dbf3ddd383b35deab1 172.18.0.6:6379@16379 master - 0 1717947605000 3 connected 10923-16383
09eb6444adf35fcb5cf185fc8a08231cd4119536 172.18.0.7:6379@16379 master - 0 1717947604000 2 connected 5461-10922
6ebd0415a63175e6e17686817dd23c49aa118e1d 172.18.0.3:6379@16379 slave d9166150899c1b584ffbcf51779751108e67c673 0 1717947605301 1 connected
624070233e88a0ce946bc76873f92c28955143e1 172.18.0.4:6379@16379 slave 09eb6444adf35fcb5cf185fc8a08231cd4119536 0 1717947604282 2 connected
d9166150899c1b584ffbcf51779751108e67c673 172.18.0.2:6379@16379 myself,master - 0 1717947602000 1 connected 0-5460
f8c8acec9a0aec8c3da2079adbb0bdc28fa85432 172.18.0.5:6379@16379 slave 692224203532603c8b1103dbf3ddd383b35deab1 0 1717947604000 3 connected

This repo also contains an example python app that connects to the cluster.

docker compose up example-app

# Output
example-app-1  | {
example-app-1  |   "172.18.0.6:6379": {
example-app-1  |     "node_id": "692224203532603c8b1103dbf3ddd383b35deab1",
example-app-1  |     "flags": "master",
example-app-1  |     "master_id": "-",
example-app-1  |     "last_ping_sent": "0",
example-app-1  |     "last_pong_rcvd": "1717947906000",
example-app-1  |     "epoch": "3",
example-app-1  |     "slots": [
example-app-1  |       [
example-app-1  |         "10923",
example-app-1  |         "16383"
example-app-1  |       ]
example-app-1  |     ],
example-app-1  |     "migrations": [],
example-app-1  |     "connected": true
example-app-1  |   },
...

TODO

This implementation is super limited. Here are some features that would be nice to have:

• Ability to reconfigure the cluster topology without taking the whole thing down
• Persistent storage using the host filesystem
• Shared configuration for number of nodes and replicas