阅读中文版本

Build an Idempotency Key Store

Idempotency Key Store — a complete walkthrough of request deduplication, covering key store, result cache, in-flight tracker, and testing.

idempotency_key_store/
├── src/index.ts
├── src/key-store.ts
├── src/result-cache.ts
├── tests/idempotency_key_store.test.ts
└── package.json

// Key store: request deduplication core
export function keystore(input: KeystoreInput): KeystoreResult {
  // Build the key store that anchors request deduplication and defines the core data shape.
  return processKeystore(input);
}

// Result cache — handles request deduplication Add the result cache so
export class Resultcache {
  constructor(private opts: ResultcacheOptions) {}
  async run(ctx: Context): Promise<ResultcacheOutput> {
    return this.execute(ctx);
  }
}

// In-flight tracker: request deduplication stage
export const inflighttracker = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// TTL evictor for request deduplication
export async function ttlevictor(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Conflict responder: request deduplication finalization
export function conflictresponder(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a request deduplication system whose reason for existing is remembering the result of a request keyed by an idempotency key so a retry returns the same response. The temptation in request deduplication is to bolt the interesting part onto a working prototype and discover the hard cases in production; this build inverts that order and treats a key collision where two unrelated requests share a key and the second receives the first request result as the first thing designed, not the last.

You will implement Key store, Result cache, In-flight tracker, TTL evictor, and Conflict responder, each as a module with one responsibility and a test that proves it in isolation. The request deduplication pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a key collision where two unrelated requests share a key and the second receives the first request result.

The walkthrough is written for a developer who has used a request deduplication system but has not had to make one honest. It favors explicit data shapes over clever abstractions, and it names the failure each design choice prevents rather than presenting the choice as a style preference.

Each phase ends with something observable — a key store that returns the right shape, a result cache that rejects bad input, an integration test that goes green — so you can stop at any point and still have a request deduplication system that works for the cases covered so far.

What you will build

The finished request deduplication system exposes five capabilities, each with a focused test:

• Key store
• Result cache
• In-flight tracker
• TTL evictor
• Conflict responder

Key store is the foundation that every later stage assumes; Result cache is where request deduplication correctness lives; In-flight tracker handles a key collision where two unrelated requests share a key and the second receives the first request result; TTL evictor makes the system observable; and Conflict responder is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the request deduplication logic in ordinary modules you can read in one sitting, and pulls in dependencies only for the parts that are genuinely tedious to hand-roll. That keeps the diff between the tutorial and a real codebase small, and keeps the failure modes visible in code rather than hidden behind a framework.

Architecture and data model

The system is organized around Key store, Result cache, and In-flight tracker, with TTL evictor tying them together. Data flows as a store mapping idempotency keys to either an in-flight marker or a cached response, expiring after a configurable window.

Each component is a module with a pure interface where practical: the same input produces the same output, and side effects are pushed to the edges. That discipline is what makes the phases independently testable, because a test for In-flight tracker does not need to stand up Key store and Result cache to exercise the interesting branch.

The data model is deliberately explicit. A store mapping idempotency keys to either an in-flight marker or a cached response, expiring after a configurable window Keeping the shape written down — in a schema, a type, or a fixture — means a change to the shape is a visible diff rather than a silent runtime shift, and it means the integration test can assert against a concrete shape instead of guessing.

Configuration and code are kept apart: the configuration says what request deduplication should do, the modules say how. A reviewer can change a threshold or a policy without reading the implementation, which keeps the risky changes small and lets the diff stand as the review.

The failure path through In-flight tracker is designed before the happy path is finished, because a key collision where two unrelated requests share a key and the second receives the first request result is the case that distinguishes a request deduplication system from a demo. Every component declares what it does when its inputs are wrong, and the integration test asserts each of those declarations.

Prerequisites and dependencies

Have these ready before you install anything, so the request deduplication setup is not blocked by a missing tool:

• Node.js 20 LTS or later
• Familiarity with request deduplication fundamentals
• A local environment to run the test suite

The runtime dependencies are chosen to keep the request deduplication datapath legible:

• redis (^4.7.0) — Key and result store
• fastify (^5.0.0) — Middleware API

# setup for idempotency-key-store
pnpm init
pnpm add redis fastify
pnpm add -D typescript tsx vitest

Versions are pinned to a major range so the request deduplication install is reproducible; commit the lockfile. The key store reads its operational parameters from the environment and validates them at startup, so a missing value fails the boot rather than the first request deduplication request.

Directory structure

Keep each request deduplication component in its own file so it can be tested without starting the others:

idempotency_key_store/
├── src/index.ts
├── src/key-store.ts
├── src/result-cache.ts
├── tests/idempotency_key_store.test.ts
└── package.json

Phase 1: Key store

Build the key store that anchors request deduplication and defines the core data shape.

// Key store: request deduplication core
export function keystore(input: KeystoreInput): KeystoreResult {
  // Build the key store that anchors request deduplication and defines the core data shape.
  return processKeystore(input);
}

After this phase the key store works in isolation and is covered by a unit test. The key store defines the data shape that every later stage assumes, so getting it right first means the later phases compile against a stable contract.

Treat the key store as the single source of truth for the request deduplication data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Result cache or In-flight tracker where its effect would be harder to localize.

Phase 2: Result cache

Add the result cache so request deduplication can remembering the result of correctly.

// Result cache — handles request deduplication Add the result cache so
export class Resultcache {
  constructor(private opts: ResultcacheOptions) {}
  async run(ctx: Context): Promise<ResultcacheOutput> {
    return this.execute(ctx);
  }
}

After this phase the result cache works in isolation and is covered by a unit test. This is where request deduplication correctness lives: the result cache is what makes remembering the result of a request keyed by an idempotency key so a retry returns the same response true, and its test is the proof.

Keep the result cache pure where you can. A function that takes an input and returns an output is trivial to test and impossible to mock wrong; a class with hidden state is the opposite. Push side effects to the caller and keep the result cache about the transformation.

Phase 3: In-flight tracker

Wire the in-flight tracker to handle a key collision where without losing data.

// In-flight tracker: request deduplication stage
export const inflighttracker = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the in-flight tracker works in isolation and is covered by a unit test. The in-flight tracker is the difference between a request deduplication system that works on the happy path and one that survives a key collision where two unrelated requests share a key and the second receives the first request result.

Design the failure before the success. Decide what in-flight tracker does when a key collision where two unrelated requests share a key and the second receives the first request result: reject, retry, shed, or degrade — and write that test first. A request deduplication system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: TTL evictor

Integrate the ttl evictor to make request deduplication observable and testable end to end.

// TTL evictor for request deduplication
export async function ttlevictor(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

After this phase the ttl evictor works in isolation and is covered by a unit test. Observability is not a finishing touch; it is how you know the request deduplication system is doing what it claims in production.

Emit one structured event per meaningful action, with enough context to localize a failure without a second round-trip. The ttl evictor is what turns “something is slow” into “the result cache is slow for inputs matching this shape,” which is the difference between an alert and an investigation.

Phase 5: Conflict responder

Finalize the conflict responder so the request deduplication system degrades gracefully under overload.

// Conflict responder: request deduplication finalization
export function conflictresponder(state: State): Result {
  validate(state);
  return finalize(state);
}

After this phase the conflict responder works in isolation and is covered by a unit test. The conflict responder is what keeps overload from becoming corruption: when the request deduplication system is asked for more than it can do, it degrades in a defined way instead of failing in an undefined one.

Pick a degradation strategy for conflict responder — shed, queue, or backpressure — and make it explicit in code and in a test. Leaving the behavior to chance is how a request deduplication traffic spike becomes a data loss incident.

Testing

Each phase has a focused test: key store, result cache, and in-flight tracker are exercised in isolation, then an integration test drives the full request deduplication pipeline and asserts the a key collision where case is handled.

Start with unit tests that pin each component’s contract: the key store returns the documented shape, the result cache transforms correctly, and the in-flight tracker fails in the documented way. These tests are fast and they localize regressions to the component that broke.

Then add an integration test that composes the components and drives a realistic request deduplication scenario end to end. Assert the happy path first, then deliberately trigger a key collision where two unrelated requests share a key and the second receives the first request result and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A request deduplication suite must also prove the negative: bad input is rejected at the boundary, overload is shed rather than queued indefinitely, and a partial failure does not leave the system in an inconsistent state. These are the tests that catch the regressions a happy-path suite never would.

The entire request deduplication suite runs with no external service; every dependency is either in-process or stubbed, so every case is reproducible in CI. That is what lets the in-flight tracker failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the request deduplication test suite
pnpm vitest run

Deployment and operations

Run at least two request deduplication instances behind a load balancer and share state in Redis. Set resource limits so request deduplication overload surfaces as backpressure rather than failure.

Run at least two stateless instances behind a load balancer, and move any shared state into a shared store so the request deduplication behavior is global rather than per instance. Size limits so the system sheds load before it runs out of memory, because a request deduplication system that OOMs under load fails in a way that is hard to reason about.

Drain on deploy: stop accepting new work, let in-flight request deduplication operations finish, then exit. Because the system holds no state that cannot be reconstructed, a rollout is a matter of cycling instances one at a time while watching the error rate.

When a dependency is unavailable, decide explicitly whether the request deduplication system fails closed or degrades. Leaving that choice to chance is how a dependency outage becomes a silent data loss, so make the choice visible in code and in a test.

Security considerations

Validate all request deduplication inputs at the boundary, reject oversized payloads before buffering them, and keep secrets in the environment rather than the repository.

The request deduplication system is a choke point, so it is the right place to reject malformed input before it reaches result cache or in-flight tracker. Validate the shape and the size at the edge, and let the inner components assume what they receive is well-formed.

Never log secrets or full payloads; log enough to localize a failure without recording data that should not be on disk. Treat any external call as untrusted and validate its response before using it, because a request deduplication system that trusts its inputs blindly is one malformed response away from corruption.

Keep dependencies pinned and run an audit in CI. A transitive dependency with a known vulnerability is still a vulnerability in your request deduplication system, and the lockfile is what makes the audit meaningful.

Observability

Emit one structured event per request deduplication operation with the inputs’ shape, the outcome, and a correlation id. Track error rate and latency per component and alert on a step change rather than a single spike.

The correlation id is the thread that ties a user-visible symptom to the request deduplication decision and to the component log entry for the same operation. Without it, an investigation starts from a timestamp and a guess.

Watch the in-flight tracker failure rate separately from the overall error rate. A jump in in-flight tracker failures usually means a key collision where two unrelated requests share a key and the second receives the first request result, not a bug in result cache, and the two have different owners and different fixes.

Chart the share of request deduplication operations that shed, queue, or succeed. A change in the mix points at the cause faster than an aggregate latency can, because each class has a different root.

Troubleshooting

A key collision where two unrelated requests share a key and the second receives the first request result. Reproduce with a minimal fixture, then add a guard test so the request deduplication system rejects the condition loudly.

Performance regresses under request deduplication load. Profile with the sampling analyzer, locate the hot path in result cache, and add a targeted benchmark to lock the regression.

Keep the request deduplication dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is key store, result cache, or in-flight tracker, instead of guessing from a timeout.

Limitations and next steps

• Single-region; multi-region replication is out of scope.
• The key store is in-process and not horizontally shared by default.
• No auth layer; assume deployment behind an edge that enforces identity.

The request deduplication system is single-region by design; extending to multi-region means making key store shareable across regions, which is a real change to the consistency model rather than a config flag. The in-flight tracker is in-process, which is correct for one instance but must move to a shared store before scaling out.

There is no authentication or authorization layer; the system assumes it is deployed behind an edge that enforces identity. Adding one is straightforward but out of scope for this build, and doing it here would distract from the request deduplication correctness that is the point.

The natural next step, once the request deduplication pipeline is trusted in production, is to add a short-TTL cache for idempotent reads and a richer ttl evictor that exports histograms rather than counters. Both are additive and neither changes the contracts this build establishes.