Skip to main content
Version: 3.0.0-alpha (Diátaxis)

How to configure Redis high availability

This guide explains how to deploy a highly available Redis cluster on Hikube. The service relies on the Spotahome Redis Operator which uses Redis Sentinel to ensure automatic failover when 3 or more replicas are configured.

Prerequisites​

  • kubectl configured to interact with the Hikube API
  • Knowledge of Redis basics (see the quick start)
  • A production environment requiring high availability

Steps​

1. Configure the manifest with 3+ replicas​

To enable high availability, configure at least 3 replicas. Redis Sentinel is automatically deployed by the Spotahome operator to orchestrate leader election and failover:

redis-ha.yaml
apiVersion: apps.cozystack.io/v1alpha1
kind: Redis
metadata:
  name: my-redis-ha
spec:
  replicas: 3
  resourcesPreset: medium
  size: 5Gi
  storageClass: replicated
  authEnabled: true
note

The storageClass: replicated ensures that persistent volumes are replicated at the storage level, protecting data against physical node loss.

2. Apply the configuration​

kubectl apply -f redis-ha.yaml

3. Verify the Redis cluster​

Wait for all pods to be ready:

# Verifier l'etat des pods Redis
kubectl get pods -l app.kubernetes.io/instance=my-redis-ha -w

Expected output:

NAME                READY   STATUS    RESTARTS   AGE
my-redis-ha-0       1/1     Running   0          3m
my-redis-ha-1       1/1     Running   0          2m
my-redis-ha-2       1/1     Running   0          1m

Also verify the Redis Sentinel status:

# Verifier les pods Sentinel
kubectl get pods -l app.kubernetes.io/component=sentinel,app.kubernetes.io/instance=my-redis-ha

4. Understanding automatic failover​

With 3 replicas, Redis Sentinel provides the following functions:

  • Failure detection: Sentinel continuously monitors the master node and replicas
  • Automatic election: if the master goes down, Sentinel elects a new master from the available replicas
  • Reconfiguration: the remaining replicas are automatically reconfigured to replicate from the new master
tip

Failover is fully automatic. No manual intervention is required. The switchover time is typically a few seconds.

5. Retrieve the password​

With authEnabled: true, a password is automatically generated and stored in a Kubernetes Secret:

# Recuperer le nom du secret
kubectl get secrets | grep my-redis-ha

# Extraire le mot de passe
kubectl get secret my-redis-ha -o jsonpath='{.data.password}' | base64 -d
warning

Always enable authEnabled: true in production. Without authentication, any application with access to the cluster network can read and write to Redis.

Verification​

Verify that the HA cluster is working correctly:

# Verifier la ressource Redis
kubectl get redis my-redis-ha

# Verifier que tous les pods sont Running
kubectl get pods -l app.kubernetes.io/instance=my-redis-ha

# Verifier les services exposes
kubectl get svc -l app.kubernetes.io/instance=my-redis-ha

Expected output:

NAME                     TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)     AGE
my-redis-ha              ClusterIP   10.96.xxx.xxx   <none>        6379/TCP    5m
my-redis-ha-sentinel     ClusterIP   10.96.xxx.xxx   <none>        26379/TCP   5m

Going further​