阅读中文版本

Build a Pull-Based Metrics Aggregator

Pull-Based Metrics Aggregator — a complete walkthrough of metrics collection, covering scrape scheduler, target manager, sample parser, and testing.

metrics_aggregator_pull/
├── src/index.ts
├── src/scrape-scheduler.ts
├── src/target-manager.ts
├── tests/metrics_aggregator_pull.test.ts
└── package.json

// Scrape scheduler: metrics collection core
export function scrapescheduler(input: ScrapeschedulerInput): ScrapeschedulerResult {
  // Build the scrape scheduler that anchors metrics collection and defines the core data shape.
  return processScrapescheduler(input);
}

// Target manager — handles metrics collection Add the target manager so
export class Targetmanager {
  constructor(private opts: TargetmanagerOptions) {}
  async run(ctx: Context): Promise<TargetmanagerOutput> {
    return this.execute(ctx);
  }
}

// Sample parser: metrics collection stage
export const sampleparser = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// Downsampler for metrics collection
export async function downsampler(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Query store: metrics collection finalization
export function querystore(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a metrics collection system whose reason for existing is scraping metrics from many endpoints on a schedule and exposing them for querying without losing samples. The temptation in metrics collection 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 target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch as the first thing designed, not the last.

You will implement Scrape scheduler, Target manager, Sample parser, Downsampler, and Query store, each as a module with one responsibility and a test that proves it in isolation. The metrics collection pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch.

The walkthrough is written for a developer who has used a metrics collection 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 scrape scheduler that returns the right shape, a target manager that rejects bad input, an integration test that goes green — so you can stop at any point and still have a metrics collection system that works for the cases covered so far.

What you will build

The finished metrics collection system exposes five capabilities, each with a focused test:

• Scrape scheduler
• Target manager
• Sample parser
• Downsampler
• Query store

Scrape scheduler is the foundation that every later stage assumes; Target manager is where metrics collection correctness lives; Sample parser handles a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch; Downsampler makes the system observable; and Query store is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the metrics collection 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 Scrape scheduler, Target manager, and Sample parser, with Downsampler tying them together. Data flows as a time-indexed store of samples keyed by metric name and label set, filled by periodic scrapes of target exposition endpoints.

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 Sample parser does not need to stand up Scrape scheduler and Target manager to exercise the interesting branch.

The data model is deliberately explicit. A time-indexed store of samples keyed by metric name and label set, filled by periodic scrapes of target exposition endpoints 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 metrics collection 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 Sample parser is designed before the happy path is finished, because a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch is the case that distinguishes a metrics collection 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 metrics collection setup is not blocked by a missing tool:

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

The runtime dependencies are chosen to keep the metrics collection datapath legible:

• fastify (^5.0.0) — Query API
• undici (^6.19.0) — Scrape fetch
• prom-client (^15.0.0) — Exposition format

# setup for metrics-aggregator-pull
pnpm init
pnpm add fastify undici prom-client
pnpm add -D typescript tsx vitest

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

Directory structure

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

metrics_aggregator_pull/
├── src/index.ts
├── src/scrape-scheduler.ts
├── src/target-manager.ts
├── tests/metrics_aggregator_pull.test.ts
└── package.json

Phase 1: Scrape scheduler

Build the scrape scheduler that anchors metrics collection and defines the core data shape.

// Scrape scheduler: metrics collection core
export function scrapescheduler(input: ScrapeschedulerInput): ScrapeschedulerResult {
  // Build the scrape scheduler that anchors metrics collection and defines the core data shape.
  return processScrapescheduler(input);
}

After this phase the scrape scheduler works in isolation and is covered by a unit test. The scrape scheduler 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 scrape scheduler as the single source of truth for the metrics collection data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Target manager or Sample parser where its effect would be harder to localize.

Phase 2: Target manager

Add the target manager so metrics collection can scraping metrics from many correctly.

// Target manager — handles metrics collection Add the target manager so
export class Targetmanager {
  constructor(private opts: TargetmanagerOptions) {}
  async run(ctx: Context): Promise<TargetmanagerOutput> {
    return this.execute(ctx);
  }
}

After this phase the target manager works in isolation and is covered by a unit test. This is where metrics collection correctness lives: the target manager is what makes scraping metrics from many endpoints on a schedule and exposing them for querying without losing samples true, and its test is the proof.

Keep the target manager 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 target manager about the transformation.

Phase 3: Sample parser

Wire the sample parser to handle a target that stalls without losing data.

// Sample parser: metrics collection stage
export const sampleparser = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the sample parser works in isolation and is covered by a unit test. The sample parser is the difference between a metrics collection system that works on the happy path and one that survives a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch.

Design the failure before the success. Decide what sample parser does when a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch: reject, retry, shed, or degrade — and write that test first. A metrics collection system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: Downsampler

Integrate the downsampler to make metrics collection observable and testable end to end.

// Downsampler for metrics collection
export async function downsampler(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

After this phase the downsampler works in isolation and is covered by a unit test. Observability is not a finishing touch; it is how you know the metrics collection 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 downsampler is what turns “something is slow” into “the target manager is slow for inputs matching this shape,” which is the difference between an alert and an investigation.

Phase 5: Query store

Finalize the query store so the metrics collection system degrades gracefully under overload.

// Query store: metrics collection finalization
export function querystore(state: State): Result {
  validate(state);
  return finalize(state);
}

After this phase the query store works in isolation and is covered by a unit test. The query store is what keeps overload from becoming corruption: when the metrics collection 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 query store — shed, queue, or backpressure — and make it explicit in code and in a test. Leaving the behavior to chance is how a metrics collection traffic spike becomes a data loss incident.

Testing

Each phase has a focused test: scrape scheduler, target manager, and sample parser are exercised in isolation, then an integration test drives the full metrics collection pipeline and asserts the a target that stalls case is handled.

Start with unit tests that pin each component’s contract: the scrape scheduler returns the documented shape, the target manager transforms correctly, and the sample parser 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 metrics collection scenario end to end. Assert the happy path first, then deliberately trigger a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A metrics collection 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 metrics collection 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 sample parser failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the metrics collection test suite
pnpm vitest run

Deployment and operations

Run at least two metrics collection instances behind a load balancer and share state in Redis. Set resource limits so metrics collection 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 metrics collection behavior is global rather than per instance. Size limits so the system sheds load before it runs out of memory, because a metrics collection 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 metrics collection 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 metrics collection 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 metrics collection inputs at the boundary, reject oversized payloads before buffering them, and keep secrets in the environment rather than the repository.

The metrics collection system is a choke point, so it is the right place to reject malformed input before it reaches target manager or sample parser. 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 metrics collection 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 metrics collection system, and the lockfile is what makes the audit meaningful.

Observability

Emit one structured event per metrics collection 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 metrics collection decision and to the component log entry for the same operation. Without it, an investigation starts from a timestamp and a guess.

Watch the sample parser failure rate separately from the overall error rate. A jump in sample parser failures usually means a target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch, not a bug in target manager, and the two have different owners and different fixes.

Chart the share of metrics collection 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 target that stalls and blocks the scrape loop, delaying samples for every other target in the same batch. Reproduce with a minimal fixture, then add a guard test so the metrics collection system rejects the condition loudly.

Performance regresses under metrics collection load. Profile with the sampling analyzer, locate the hot path in target manager, and add a targeted benchmark to lock the regression.

Keep the metrics collection dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is scrape scheduler, target manager, or sample parser, instead of guessing from a timeout.

Limitations and next steps

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

The metrics collection system is single-region by design; extending to multi-region means making scrape scheduler shareable across regions, which is a real change to the consistency model rather than a config flag. The sample parser 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 metrics collection correctness that is the point.

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