05Design patternsIdempotency and Exactly-Once Semantics
05 · Reusable pattern
Idempotency and Exactly-Once Semantics
Make retries safe with stable operation keys, atomic state transitions, and deduplicated side effects.Lesson spine
What you need to understand.
Exactly-once effects are usually built from at-least-once delivery plus an idempotent, atomic state transition.
Idempotency key
Scope a stable client key to an operation and principal, store the request fingerprint, and return the original result on retry.
Atomic claim
Create the dedupe record and business effect in one transaction or use a conditional state transition that only one attempt can win.
Delivery semantics
At-most-once risks loss; at-least-once risks duplicates; exactly-once processing depends on the effect boundary, not broker marketing.
Database techniques
Use unique constraints, upserts, compare-and-set, or a processed-message table inside the same transaction.
Transactional outbox
Commit business state and an outgoing event together, then publish the outbox at least once to idempotent consumers.
Retention and misuse
Keep keys for the maximum retry horizon and reject the same key when the request payload changes.
Before the boxes
Frame the decision.
What must work
Make retries safe with stable operation keys, atomic state transitions, and deduplicated side effects.
What changes the design
Key scope · retry horizon · record volume · atomicity · response size
What owns the truth
Identify the component that commits authoritative state, then separate synchronous confirmation from derived work.
What stays simple
Do not add global coordination, multi-region writes, or a specialized store until a requirement earns the complexity.
Architecture map
Trace ownership, not just traffic.
Follow the decision from left to right. Every arrow should have a reason.
Generates operation key
Validates request fingerprint
Claims key
Commits effect and response
Returns stored result
Reclaims old keys
Walk one representative request across every arrow. Say whether the handoff is synchronous or asynchronous, what identity makes a retry safe, and which step changes authoritative state.
- 01
Client — Generates operation key Define the output contract before moving to the next owner.
- 02
API — Validates request fingerprint Define the output contract before moving to the next owner.
- 03
Dedupe store — Claims key Define the output contract before moving to the next owner.
- 04
Transaction — Commits effect and response Define the output contract before moving to the next owner.
- 05
Retry — Returns stored result Define the output contract before moving to the next owner.
- 06
Expiry — Reclaims old keys Confirm the result and emit the evidence needed to reconcile it.
Decision table
Make the trade-offs explicit.
| Decision | Defensible position | Cost to acknowledge |
|---|---|---|
| Primary mechanism | Store the response beside the effect when retries must be identical. | The stronger guarantee usually adds coordination, latency, state, or operational work. |
| Sync vs. async | Keep only correctness-critical confirmation synchronous. Move derived views, notifications, analytics, and cleanup behind a durable boundary. | Async work needs idempotency, lag monitoring, replay, and a product definition for partial completion. |
| Simple vs. scaled | Begin with one logical owner and a clear API. Partition or replicate only the resource proven to be the first bottleneck. | Migration requires stable identities, versioned contracts, backfill, and a rollback path. |
Failure review
Design the recovery path.
Topic-specific risk
Reusing a key with new input or expiring too early creates duplicate effects.
ResponsePersist enough identity and state to distinguish retry, resume, compensation, and operator repair.
Dependency timeout
A timeout is ambiguous: the remote side may have failed, succeeded, or still be running.
ResponseUse deadlines, bounded backoff with jitter, idempotency keys, and a status or reconciliation path.
Overload or skew
Average capacity can look healthy while a tenant, key, partition, region, or expensive request saturates one owner.
ResponseExpose queue depth and hot-key share, apply backpressure, isolate tenants, and degrade optional work before correctness.
Evidence + level bar
Prove the design can be operated.
Health of the promise
Measure user-visible latency or freshness, correctness drift, saturation, retry volume, and time to recover. Alert on the failed promise—not only CPU.
Complete and clear
Finish the happy path, identify the state owner, choose reasonable building blocks, and explain one scale mechanism.
Trade-offs and failure
Separate read and write paths, define consistency, explain partitioning, and make duplicate or partial failure safe.
Evolution and operations
Discuss multi-region boundaries, migration, tenant isolation, capacity, observability, and how the architecture changes over time.
Interview language
Open the deep dive with a claim.
“For Idempotency and Exactly-Once Semantics, the decision I want to make explicit is this: Store the response beside the effect when retries must be identical. I’ll trace the state-changing path first, show where the result becomes durable, then test the design against the highest-risk failure and our target scale.”
08 · Retrieval check
Can you defend it without the page?
- For Idempotency and Exactly-Once Semantics, where is the correctness boundary and which failure would you test first?
- Which component owns committed truth, and what event or response proves the commit?
- Where is the first scaling or coordination bottleneck under the stated envelope?
- What happens after an ambiguous timeout or duplicate operation?
- Which complexity would you remove at one hundredth of the scale?