Redis High‑Availability Architecture and Best Practices

Introduction

Redis is an open‑source, in‑memory key‑value store written in ANSI C that supports persistence and networking. Its design prioritises performance, making it one of the fastest NoSQL databases, while high availability is essential for 24/7 internet services.

Sentinel Principle

Redis Sentinel discovers masters via configuration files, monitors them by sending periodic INFO commands, and exchanges HELLO messages with other Sentinels to announce its presence. It uses PING to detect instance health, and a failover is triggered only after a quorum of Sentinels authorises it. The elected slave becomes the new master based on priority, replication offset, and process ID, after which Sentinels broadcast the updated configuration version.

High‑Availability Architectures

Several common Redis HA solutions are presented:

Sentinel cluster + internal DNS + custom scripts.

Sentinel cluster + VIP + custom scripts.

Client‑direct connection to a Sentinel port, then to the current master (e.g., JedisSentinelPool for Java, PHP wrappers).

Sentinel cluster + Keepalived/Haproxy.

Native master/slave with Keepalived (M/S + Keepalived).

Redis Cluster (sharding with 16384 slots, gossip protocol).

Twemproxy (proxy layer with consistent hashing).

Codis (ZooKeeper‑based routing, Codis‑Proxy, Codis‑Redis).

Each architecture is accompanied by concise pros and cons, such as sub‑second failover, transparency to applications, maintenance cost, dependency on DNS or VIP, potential split‑brain, and performance impact of proxies.

Best Practices

The author recommends primarily using Sentinel + internal DNS + custom scripts or Sentinel + VIP + custom scripts. Operational tips include deploying at least five Sentinel nodes, grouping business services under a single Sentinel cluster, defining clear port ranges, implementing scripts in Python with Paramiko for SSH reuse, disabling DNS and GSSAPI in SSH, and ensuring automatic or manual failover completes within 15 seconds.

Conclusion

The presentation emphasizes the necessity of Redis high availability, summarizes Sentinel fundamentals, compares practical HA architectures, and shares field‑tested best practices to help engineers design robust Redis deployments.