Concept questionMedium~7 min

Interview question

A popular cache key expires and many requests hit the database at once. How would you prevent that?

Prevent a cache stampede on hot keys

Explains what happens when a hot key expires and how request coalescing, TTL jitter, background refresh, and stale serving help.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

A cache stampede happens when many requests miss the same key at once and all try to rebuild it. I would prevent that by making only one request rebuild the value while others wait briefly, use a stale value, or receive a bounded fallback. Adding TTL jitter also helps avoid many hot keys expiring at the same time.

Full answer below

Practice focus

Hot keys, cache stampede, request coalescing, TTL jitter, stale-while-revalidate, background refresh, and source protection.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Why this question matters
Cache stampedes are a classic production failure: the cache works until it suddenly sends all traffic back to the source at the same moment.
A cache stampede happens when many requests miss the same key at once and all try to rebuild it. I would prevent that by making only one request rebuild the value while others wait briefly, use a stale value, or receive a bounded fallback. Adding TTL jitter also helps avoid many hot keys expiring at the same time.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
Inside one process I can use single-flight request coalescing so concurrent callers share one rebuild task. Across replicas, a short lease can nominate one refresher while others serve an acceptable stale value or wait for a bounded period. TTL jitter prevents unrelated hot keys from expiring together, and source concurrency limits remain in place if coordination fails. Negative caching can protect repeated not-found lookups, but only with a short lifetime that will not hide newly created data.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
In production I design for the refresher crashing or exceeding its lease. Lock ownership is verified before release, stale serving has a hard maximum age, and callers eventually fall back or fail rather than waiting without bound. Cold starts, bulk invalidation, and region recovery are tested because they can create a stampede without normal TTL expiry. Metrics cover coalesced waiters, rebuild duration, lock contention, stale serves, miss bursts, source concurrency, and refresh failures so the protection can be proven under load.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Letting every miss rebuild the same expensive value.
Using a distributed lock without expiration or failure handling.
Setting identical TTLs on many hot keys.
Serving stale data where immediate correctness is required.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Cannot explain why the database is hit all at once.
Only suggests increasing Redis size.
No concurrency limit on the rebuild path.
No metrics for miss bursts or refresh failures.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

When is stale-while-revalidate acceptable?
How would you implement request coalescing?
What can go wrong with a Redis lock here?

Continue practice

Move to the linked follow-up, next path step, prerequisite, or deeper variant.

1 linked

Go deeper

Practice the next layer of the same subject.

Execution Plan Diagnosis

Concept questionMedium~7 min

Interview question

A popular cache key expires and many requests hit the database at once. How would you prevent that?

Prevent a cache stampede on hot keys

Explains what happens when a hot key expires and how request coalescing, TTL jitter, background refresh, and stale serving help.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

Full answer below

Practice focus

Hot keys, cache stampede, request coalescing, TTL jitter, stale-while-revalidate, background refresh, and source protection.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Why this question matters
Cache stampedes are a classic production failure: the cache works until it suddenly sends all traffic back to the source at the same moment.
A cache stampede happens when many requests miss the same key at once and all try to rebuild it. I would prevent that by making only one request rebuild the value while others wait briefly, use a stale value, or receive a bounded fallback. Adding TTL jitter also helps avoid many hot keys expiring at the same time.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
Inside one process I can use single-flight request coalescing so concurrent callers share one rebuild task. Across replicas, a short lease can nominate one refresher while others serve an acceptable stale value or wait for a bounded period. TTL jitter prevents unrelated hot keys from expiring together, and source concurrency limits remain in place if coordination fails. Negative caching can protect repeated not-found lookups, but only with a short lifetime that will not hide newly created data.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
In production I design for the refresher crashing or exceeding its lease. Lock ownership is verified before release, stale serving has a hard maximum age, and callers eventually fall back or fail rather than waiting without bound. Cold starts, bulk invalidation, and region recovery are tested because they can create a stampede without normal TTL expiry. Metrics cover coalesced waiters, rebuild duration, lock contention, stale serves, miss bursts, source concurrency, and refresh failures so the protection can be proven under load.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Letting every miss rebuild the same expensive value.
Using a distributed lock without expiration or failure handling.
Setting identical TTLs on many hot keys.
Serving stale data where immediate correctness is required.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Cannot explain why the database is hit all at once.
Only suggests increasing Redis size.
No concurrency limit on the rebuild path.
No metrics for miss bursts or refresh failures.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

When is stale-while-revalidate acceptable?
How would you implement request coalescing?
What can go wrong with a Redis lock here?

Continue practice

Move to the linked follow-up, next path step, prerequisite, or deeper variant.

1 linked

Go deeper

Practice the next layer of the same subject.

Execution Plan Diagnosis

A popular cache key expires and many requests hit the database at once. How would you prevent that?

Practice focus

Answer path

Why this question matters

Review signals

Follow-up prompts

Continue practice

Go deeper

A popular cache key expires and many requests hit the database at once. How would you prevent that?

Practice focus

Answer path

Why this question matters

Review signals

Follow-up prompts

Continue practice

Go deeper