Why Redis Cluster Limits You to DB 0 (and How to Work Around It)

Introduction

Many engineers assume they can switch databases in Redis Cluster with SELECT, but in practice only DB 0 is usable. The article explains the design reasons behind this limitation.

Reason 1: Slot‑based sharding conflicts with multiple DBs

Redis Cluster maintains a single slot map of 0‑16383. Each key is hashed with CRC16 and assigned to a slot: slot = CRC16(key) % 16384. Slots guarantee that a given key always resides on the same node.

If multiple databases were allowed, the same key could appear in different DBs and map to different slots, destroying the global slot mapping. The author cites Antirez: "Redis Cluster never supported multiple databases. Only DB 0 is allowed."

Reason 2: Slot migration would become impossible

Cluster operations such as scaling out or shrinking involve migrating slots, e.g., migrate slot 5500 from node A → node B. With multiple DBs, each DB would need its own slot metadata and migration logic, causing exponential complexity and breaking consistency.

Why the design chooses a single DB

The architecture emphasizes distributed consistency, not logical isolation. A DB index is merely a logical namespace, while the cluster’s goals are scalability, fault‑tolerance, and data consistency. Supporting multiple DBs would add unnecessary overhead and weaken isolation for large‑scale systems.

How Java projects can simulate multiple databases

Key‑prefix isolation (most common) : prepend a logical prefix to keys, e.g., prod:user:1001, test:user:1001, appA:config:*. RedisTemplate can add the prefix automatically. Advantages: simple; Drawbacks: requires managing prefix conventions.

Multiple Redis instances : run separate Redis processes for different domains (user cache, order cache, session store). Provides physical isolation.

Multiple Redis Clusters : in micro‑service architectures, allocate a dedicated cluster per domain (e.g., user‑redis‑cluster, search‑redis‑cluster, trade‑redis‑cluster) to achieve resource, fault, and scaling isolation.

Can you SELECT another DB in Redis Cluster?

The answer is no. The cluster forces DB 0 because the 16,384‑slot sharding would be corrupted by additional DBs, making slot migration and data distribution impossible. The source code explicitly disables multi‑DB support.

Standard answer (one‑minute summary)

Redis Cluster uses a slot‑based sharding mechanism (16384 slots) where every key’s CRC16 hash determines its slot.

Multiple DBs would break the slot model, causing the same key to map to different slots across DBs.

Slot migration would become extremely complex or unmaintainable.

Therefore the Redis source code hard‑codes support for only DB 0 in cluster mode.

When logical isolation is needed, use key prefixes, separate instances, or separate clusters instead of Redis’s DB concept.