阅读中文版本

Build a Distributed Tracing Collector

Distributed Tracing Collector — a complete walkthrough of trace collection, covering span ingest, trace assembler, span store, and testing.

distributed_tracing_collector/
├── src/index.ts
├── src/span-ingest.ts
├── src/trace-assembler.ts
├── tests/distributed_tracing_collector.test.ts
└── package.json

// Span ingest: trace collection core
export function spaningest(input: SpaningestInput): SpaningestResult {
  // Build the span ingest that anchors trace collection and defines the core data shape.
  return processSpaningest(input);
}

// Trace assembler — handles trace collection Add the trace assembler so
export class Traceassembler {
  constructor(private opts: TraceassemblerOptions) {}
  async run(ctx: Context): Promise<TraceassemblerOutput> {
    return this.execute(ctx);
  }
}

// Span store: trace collection stage
export const spanstore = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

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

// Tail-based sampler: trace collection finalization
export function tailbasedsampler(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a trace collection system whose reason for existing is accepting spans from many services and assembling them into traces with correct causal order. The temptation in trace 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 dropped span that leaves a trace with a broken parent link and an incomplete call tree as the first thing designed, not the last.

You will implement Span ingest, Trace assembler, Span store, Sampling policy, and Tail-based sampler, each as a module with one responsibility and a test that proves it in isolation. The trace collection pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a dropped span that leaves a trace with a broken parent link and an incomplete call tree.

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

What you will build

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

• Span ingest
• Trace assembler
• Span store
• Sampling policy
• Tail-based sampler

Span ingest is the foundation that every later stage assumes; Trace assembler is where trace collection correctness lives; Span store handles a dropped span that leaves a trace with a broken parent link and an incomplete call tree; Sampling policy makes the system observable; and Tail-based sampler is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the trace 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 Span ingest, Trace assembler, and Span store, with Sampling policy tying them together. Data flows as a store of spans keyed by trace id, assembled into a tree by parent span references and retained for 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 Span store does not need to stand up Span ingest and Trace assembler to exercise the interesting branch.

The data model is deliberately explicit. A store of spans keyed by trace id, assembled into a tree by parent span references and retained for 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 trace 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 Span store is designed before the happy path is finished, because a dropped span that leaves a trace with a broken parent link and an incomplete call tree is the case that distinguishes a trace 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 trace collection setup is not blocked by a missing tool:

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

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

• fastify (^5.0.0) — Ingest API
• opentelemetry (^1.27.0) — Span format
• level (^8.0.0) — Span store

# setup for distributed-tracing-collector
pnpm init
pnpm add fastify opentelemetry level
pnpm add -D typescript tsx vitest

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

Directory structure

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

distributed_tracing_collector/
├── src/index.ts
├── src/span-ingest.ts
├── src/trace-assembler.ts
├── tests/distributed_tracing_collector.test.ts
└── package.json

Phase 1: Span ingest

Build the span ingest that anchors trace collection and defines the core data shape.

// Span ingest: trace collection core
export function spaningest(input: SpaningestInput): SpaningestResult {
  // Build the span ingest that anchors trace collection and defines the core data shape.
  return processSpaningest(input);
}

After this phase the span ingest works in isolation and is covered by a unit test. The span ingest 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 span ingest as the single source of truth for the trace collection data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Trace assembler or Span store where its effect would be harder to localize.

Phase 2: Trace assembler

Add the trace assembler so trace collection can accepting spans from many correctly.

// Trace assembler — handles trace collection Add the trace assembler so
export class Traceassembler {
  constructor(private opts: TraceassemblerOptions) {}
  async run(ctx: Context): Promise<TraceassemblerOutput> {
    return this.execute(ctx);
  }
}

After this phase the trace assembler works in isolation and is covered by a unit test. This is where trace collection correctness lives: the trace assembler is what makes accepting spans from many services and assembling them into traces with correct causal order true, and its test is the proof.

Keep the trace assembler 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 trace assembler about the transformation.

Phase 3: Span store

Wire the span store to handle a dropped span that without losing data.

// Span store: trace collection stage
export const spanstore = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the span store works in isolation and is covered by a unit test. The span store is the difference between a trace collection system that works on the happy path and one that survives a dropped span that leaves a trace with a broken parent link and an incomplete call tree.

Design the failure before the success. Decide what span store does when a dropped span that leaves a trace with a broken parent link and an incomplete call tree: reject, retry, shed, or degrade — and write that test first. A trace collection system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: Sampling policy

Integrate the sampling policy to make trace collection observable and testable end to end.

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

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

Phase 5: Tail-based sampler

Finalize the tail-based sampler so the trace collection system degrades gracefully under overload.

// Tail-based sampler: trace collection finalization
export function tailbasedsampler(state: State): Result {
  validate(state);
  return finalize(state);
}

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

Testing

Each phase has a focused test: span ingest, trace assembler, and span store are exercised in isolation, then an integration test drives the full trace collection pipeline and asserts the a dropped span that case is handled.

Start with unit tests that pin each component’s contract: the span ingest returns the documented shape, the trace assembler transforms correctly, and the span store 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 trace collection scenario end to end. Assert the happy path first, then deliberately trigger a dropped span that leaves a trace with a broken parent link and an incomplete call tree and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A trace 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 trace 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 span store failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the trace collection test suite
pnpm vitest run

Deployment and operations

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

The trace collection system is a choke point, so it is the right place to reject malformed input before it reaches trace assembler or span store. 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 trace 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 trace collection system, and the lockfile is what makes the audit meaningful.

Observability

Emit one structured event per trace 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 trace 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 span store failure rate separately from the overall error rate. A jump in span store failures usually means a dropped span that leaves a trace with a broken parent link and an incomplete call tree, not a bug in trace assembler, and the two have different owners and different fixes.

Chart the share of trace 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 dropped span that leaves a trace with a broken parent link and an incomplete call tree. Reproduce with a minimal fixture, then add a guard test so the trace collection system rejects the condition loudly.

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

Keep the trace collection dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is span ingest, trace assembler, or span store, instead of guessing from a timeout.

Limitations and next steps

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

The trace collection system is single-region by design; extending to multi-region means making span ingest shareable across regions, which is a real change to the consistency model rather than a config flag. The span store 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 trace collection correctness that is the point.

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