RedisVisualFault simulator
Stop 8 · Fault simulator
Break Redis on purpose and watch the fallout live in hit rate and latency — breakdown, avalanche, hot keys, TTL, eviction.
SIM
The pitfalls you read about — now you can watch them happen
Below is a simulation that is actually running: each second the engine generates your chosen QPS of requests against a mock cache with and . A hit is fast (~1ms); a miss falls through to the “database” (~40ms, and slower once more misses arrive at once than the database can handle). Move the controls on the left and press the fault buttons; the hit rate, latency, and DB load on the right react live. Flip the two fixes (TTL jitter / single-flight) to watch the curves recover.
Control panel
Fixes
Trigger a fault
100%Hit rate
1msAvg latency
1msp99 latency
0/sDB load
0 / 160Cache used
Hit rate (higher is better)0%
p99 latency (lower is better)0ms
Keys in the cache (first 60)hotfreshexpiringempty / evicted
One-click scenarios
Event log
No events yet — change a control or trigger a fault.
What the curves teach
- Hit rate is set by capacity, TTL and access pattern together; once capacity < keyspace, LRU eviction is unavoidable and the hit rate caps out.
- Latency climbs steeply: once rebuilds exceed what the DB can handle per tick, latency grows with the excess (linearly with the overload in this model) — that is what makes breakdown and avalanche dangerous.
- Single-flight fixes breakdown (collapses N rebuilds into 1); TTL jitter fixes avalanche (de-synchronizes expiry). Right fix, right fault — one toggle shows it.