What causes Redis latency spikes, and how do you diagnose them?

0-2 years experience

1. 1Your app started timing out occasionally when writing to Redis — what three commands or tools would you run first to figure out why?
2. 2If Redis suddenly starts taking 500ms for simple GETs, but the CPU is fine, what’s the most likely culprit and how would you check?
3. 3A teammate says Redis is slow because it’s using too much memory — how would you verify that’s actually the problem?

2-5 years experience

1. 1Our order service started seeing 2-second latency spikes every 10 minutes — the slowlog shows no slow commands, but disk I/O is spiking. What’s going on and how do you fix it?
2. 2We migrated Redis to a new VM and latency doubled. The network ping is the same, memory is fine. What three things would you investigate and why?
3. 3A feature that uses EVAL scripts started timing out under load — how do you determine if it’s the script, Redis itself, or the client?

5-8 years experience

1. 1Our Redis cluster has 100+ instances and latency spikes correlate with memory fragmentation — how would you design a monitoring and mitigation strategy that doesn’t require restarts?
2. 2We’re seeing periodic 100ms+ latency spikes during AOF rewrite, even with no client traffic. How would you redesign the persistence strategy to eliminate this without losing durability?
3. 3A critical service uses Redis for session storage and sees 500ms spikes every time a large key is deleted. How do you architect around this without changing the data model?

8+ years experience

1. 1We’re migrating from a single-node Redis to a clustered setup, but latency spikes during shard rebalancing are breaking SLAs — how do you design a zero-downtime migration with predictable performance?
2. 2Our Redis instances are running on shared cloud VMs and latency spikes correlate with noisy neighbors — what architectural changes would you push for to guarantee consistent performance at scale?
3. 3A legacy system relies on Redis for real-time leaderboards with 10M+ keys, and fragmentation causes daily restarts. How do you evolve this architecture to eliminate operational fragility without a full rewrite?