Understanding Redis Hotkeys: Issues, Detection Methods, and Mitigation Strategies

1 What is a Hotkey and what problems does it cause?

1.1 Definition

A Hotkey in Redis refers to a key that receives a disproportionately high number of read/write requests, becoming a hotspot.

1.2 Problems

1.2.1 Network Issues

When a single node cannot offload traffic, the hotkey saturates the instance’s network interface, leading to latency.

1.2.2 Cache Penetration

If the hotkey expires or its node fails, requests fall back to the underlying database.

1.2.3 Master‑Slave Replication Lag

Asynchronous replication can be delayed or interrupted when a hotkey continuously consumes bandwidth.

2 How to Detect Hotkeys

2.1 Client‑Side Statistics

Collect key access metrics on the business client or proxy layer.

Advantages

Low implementation cost.

Flexible configuration of metrics and caching logic.

Disadvantages

Requires instrumentation in business code, increasing complexity.

Aggregation introduces monitoring latency.

2.2 Redis MONITOR Command

The MONITOR command streams every command executed by Redis, allowing real‑time analysis of hotkey activity.

Recommendation

Facebook’s open‑source redis-faina (Python) provides analysis tools for MONITOR output.

Advantages

Shows read vs write behavior.

Accurately identifies client source.

Disadvantages

The MONITOR command itself can reduce throughput by more than 50% in high‑load environments.

In this particular case, running a single MONITOR client can reduce the throughput by more than 50%. Running more MONITOR clients will reduce throughput even more

2.3 Redis HOTKEYS Command (since 4.0.3)

The hotkeys command lists the most frequently accessed keys, useful for low‑traffic scenarios after configuring an eviction policy.

Advantages

Easy to use – built‑in command.

Provides real‑time hotspot information.

Disadvantages

Performance impact on large datasets.

Results depend on internal sampling; may not be 100% accurate.

Only basic information is returned; deeper analysis requires application‑side data.

2.4 TCP Packet Capture

Capturing Redis traffic with tools like Elastic’s packetbeat offers non‑intrusive, performance‑neutral insight.

Advantages

Real‑time visibility of client‑server communication.

Works for any Redis deployment.

Independent of business code or Redis configuration.

Disadvantages

Higher implementation complexity and cost.

Potential instability in unreliable networks.

May expose sensitive data in packets.

3 Mitigation Strategies

3.1 Redis Cluster Sharding

Distribute keys across multiple nodes to balance load; suitable for general workloads but less effective for isolated hotkeys.

3.2 Multi‑Level Caching

Combine local, distributed, global, and regional caches to alleviate pressure on hotkeys.

3.3 Monitoring Optimisation

Set up alerts on QPS, memory, and network metrics; automate scaling or slot migration when thresholds are breached.

3.4 Split Business Keys

When feasible, break a hotkey into sub‑keys to spread traffic.

3.5 Cache TTL and Eviction Policy

Configure appropriate TTLs and use LRU/LFU policies to evict cold data while retaining hot data, preventing cache stampede.

References

[1] redis-faina: https://github.com/facebookarchive/redis-faina

[2] MONITOR command: https://redis.io/commands/monitor/

[3] packetbeat: https://www.elastic.co/cn/beats/packetbeat