Redis Cache Deployment with EmptyDir Volume

Redis Cache on Kubernetes

Redis as a cache does not need persistent storage — if the pod restarts, the cache can be rebuilt. However, using an emptyDir volume for Redis's RDB dumps provides faster restarts when the dump file exists.

Key Configuration

apiVersion: apps/v1
kind: Deployment
metadata:
  name: redis
  labels:
    app: "redis"
    role: "cache"
spec:
  replicas: 1
  template:
    spec:
      containers:
        - name: redis
          image: redis:7-alpine
          ports:
            - name: redis
              containerPort: 6379
          resources:
            requests:
              cpu: "100m"
              memory: "128Mi"
            limits:
              cpu: "250m"
              memory: "256Mi"
          volumeMounts:
            - name: redis-data
              mountPath: /data
          livenessProbe:
            tcpSocket:
              port: 6379
            initialDelaySeconds: 10
            periodSeconds: 10
          readinessProbe:
            tcpSocket:
              port: 6379
            initialDelaySeconds: 5
            periodSeconds: 5
      volumes:
        - name: redis-data
          emptyDir: {}

EmptyDir vs PVC

Feature	emptyDir	PVC
Persists across container restarts	Yes	Yes
Persists across pod restarts	No	Yes
Performance	RAM-speed (tmpfs) or disk	Depends on storage class
Use case	Cache, temp data	Persistent queue, session store

For Redis as a pure cache, emptyDir is appropriate. If you need data to survive pod rescheduling (e.g., Redis as a session store), use a PVC instead.

TCP Socket Probe

Redis does not have an HTTP endpoint by default, so we use a TCP socket probe on port 6379. This checks that the Redis process is listening and accepting connections. For more thorough checks, you could use an exec probe running redis-cli ping.

Memory Management

Set the Redis maxmemory directive to slightly less than the container memory limit (e.g., 200Mi if the limit is 256Mi) and configure an eviction policy like allkeys-lru. This prevents Redis from being OOM-killed by the container runtime.