Role practice · the whole loop
Backend developer interview questions: APIs, data and failure
Practice backend developer interviews on what breaks in production: retries, slow queries, race conditions, migrations and downstream failures, and how you would know.
Free during early access. Practice runs in the app, which needs an account.
Prep sheetBackend developer
- 01API design
- 02Databases and indexing
- 03Concurrency
- 04Failure handling
01Common prompts
Backend developer interview questions by round
These backend interview questions are practice prompts written or adapted by InterviewWare, not questions reported by any employer. They are grouped by round, with the signals an interviewer listens for and the follow-up that usually comes next.
| Question | What the interviewer listens for | Likely follow-up |
|---|---|---|
| R1 · Background & motivation | ||
| Q01 Walk me through the data model of a service you built, and one thing you would change. | The main entities and access patterns, the index or key choices that follow from them, and an honest change with its migration cost. | Which query would break first if traffic grew tenfold? |
| R2 · Role depth & case | ||
| Q02 How do you make an API operation safe to retry? | Idempotency keys stored with the result, naturally idempotent operations where possible, and what the client sees on a duplicate. | Two identical requests reach different servers at the same moment. What stops both from running? |
| Q03 How would you investigate a slow database query? | Reproducing with real parameters, reading the query plan, checking index use and row estimates, and weighing a new index against its write cost. | The query is fast in staging and slow in production. What differs? |
| Q04 What should happen when a downstream service becomes unavailable? | Timeouts, bounded retries with backoff and jitter, circuit breaking, and a deliberate degraded response instead of a cascading failure. | Your retries made the outage worse. Why, and what would you change? |
| Q05 Two customers try to buy the last item in stock at the same moment. How do you prevent overselling? | An atomic conditional update or a row lock, optimistic concurrency with a version number, or a reservation that expires, and the trade-off of each under load. | During a flash sale the hot row becomes a bottleneck. What now? |
| Q06 Add a column to a large, busy table and backfill it without downtime. | Expand-and-contract steps, a throttled backfill in batches, code that tolerates both schemas, and a rollback point. | The backfill is halfway done and replication lag spikes. What do you do? |
| Q07 When would you process work through a queue instead of inside the request? | Latency and failure isolation as the reasons; at-least-once delivery with idempotent consumers; how the user learns the result. | A poison message crashes the consumer on every attempt. How do you stop it blocking the queue? |
| Q08 A popular cache key expires and the database is flooded. How do you prevent it? | Coalescing requests or locking around recomputation, staggered or early expiry, briefly serving stale data, and watching the hit rate. | How would you decide how stale the data is allowed to be? |
| Q09 After a deploy, p99 latency doubled but the median didn’t move. Where do you look? | Causes specific to the tail — a slow dependency on some requests, lock contention, GC pauses, cold caches on new instances — and comparing traces of slow and normal requests. | Would you roll back first or investigate first? Why? |
| Q10 Server-side sessions or signed tokens for authenticating API clients: how do you choose? | Revocation and expiry, where state lives, token size, and rotating signing keys. | A token leaks. How quickly can your design shut it off? |
| R3 · Judgment & collaboration | ||
| Q11 A product manager wants to skip the load test to hit a launch date. What do you say? | Putting a size on the risk, offering a smaller test or a staged rollout, and agreeing who owns the decision. | You are overruled and the launch goes ahead. What do you put in place? |
02Try one question
Answer a backend interview question
Draft an answer to one of the questions above. It stays in this browser until you ask for feedback.
Practice question · Backend developer
How do you make an API operation safe to retry?
Feedback needs a free account (one step, no card). Your draft is kept only in this browser and filled in after you sign up; nothing is sent until you submit it.
03Worked example
Our payment webhook sometimes arrives twice. How would you handle it?
Mid-level answer outline, about two minutes. Illustrative sample written by InterviewWare, not a real candidate’s answer.
- 01
Assume
I’ll assume the provider sends an event ID and retries until we return a success status.
- 02
Design
Store each event ID under a unique constraint in the same transaction as the order update, so a duplicate insert fails and we return success without reprocessing.
- 03
Failure
If we crash after the commit but before responding, the retry hits the unique constraint and is acknowledged safely.
- 04
Observe
I’d count duplicates per provider, so a sudden rise tells us their retries or our responses have changed.
04Coaching notes
How to answer backend interview questions
What works
- T1Describe the failure path as carefully as the happy path.
- T2State consistency and latency requirements before choosing storage or a cache.
- T3Name the metric or log line that would tell you something is wrong.
- T4Give every migration or rollout answer a rollback step.
Common mistakes in backend interviews
- M1Saying “add retries” without backoff, limits or idempotency.
- M2Reaching for a distributed lock where a conditional update would do.
- M3Proposing an index without mentioning its write and storage cost.
- M4Designing only for average latency and forgetting the tail.
- M5Claiming exactly-once delivery without explaining how it is achieved.
05Tools that help
Tools for backend interview prep
- JS · PythonPublic tests
Technical interview lab
Sketch the retry and idempotency path on the system-design whiteboard before you explain it.
- Realtime15–60 min
AI voice mock interview
Walk through an incident out loud in a voice session, where the interviewer can cut in with “how did you know?”
- 4 interviewersR1–R3Text · voice
Conversational AI mock interviews
Choose the engineering-lead interviewer, who pushes for latency, error rates and failure modes.
06Company prep
Company notes for backend interviews
Summarised from each company’s public hiring pages. The drills are InterviewWare practice prompts, not questions these companies use, and InterviewWare is not affiliated with any of them.
All company prep sourcesAmazon
Amazon’s SDE II interview prep page asks candidates to discuss reliability and scale explicitly and to test edge cases.
Practice drill
Design an order-processing workflow that tolerates duplicate events and partial failures.
Practice this drillOfficial sources: Amazon · SDE II Interview Prep
Sources checked September 27, 2026
Microsoft
Microsoft’s technical-interviewing tips ask you to test normal, boundary and failure cases.
Practice drill
How would you test a file-upload service that must recover from interrupted transfers?
Practice this drillOfficial sources: Microsoft · Technical interviewing
Sources checked September 27, 2026
More directions
Explore another direction
Early access · free
Log your first rep today.
Bring a resume or a job description — or just the role you want. The log starts with your first answer.