Abstract
- In distributed systems like Kubernetes, we can have app instances running on multiple machines. In such cases, we can deploy a Redis cluster, and all the app instances read/write from the same cluster
Managed Redis
Managed Redis makes cluster manegement easier!
On GCP, Memorystore for Redis is the usual pick. It integrates well with GKE over private VPC. Ops (patching, monitoring, failover) are handled for you. If you need sharding or persistence, use Memorystore Redis Cluster.
Handling failures and persistance
Replicas provide high availability, but replication is async, so you may lose some writes on failover.
For durability, enable persistence (AOF/RDB) or scheduled backups. In GCP, only Memorystore Redis Cluster supports managed persistence.