阅读中文版本

Build a Load Balancer with Health Checks

Load Balancer with Health Checks — a complete walkthrough of traffic distribution, covering upstream pool, active health probe, passive failure counter, and testing.

load_balancer_health/
├── src/index.ts
├── src/upstream-pool.ts
├── src/active-health-probe.ts
├── tests/load_balancer_health.test.ts
└── package.json

// Upstream pool: traffic distribution core
export function upstreampool(input: UpstreampoolInput): UpstreampoolResult {
  // Build the upstream pool that anchors traffic distribution and defines the core data shape.
  return processUpstreampool(input);
}

// Active health probe — handles traffic distribution Add the active health probe
export class Activehealthprobe {
  constructor(private opts: ActivehealthprobeOptions) {}
  async run(ctx: Context): Promise<ActivehealthprobeOutput> {
    return this.execute(ctx);
  }
}

// Passive failure counter: traffic distribution stage
export const passivefailurecounter = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// Ejection policy for traffic distribution
export async function ejectionpolicy(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Pick algorithm: traffic distribution finalization
export function pickalgorithm(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a traffic distribution system whose reason for existing is spreading requests across upstreams and ejecting unhealthy ones before clients see failures. The temptation in traffic distribution 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 health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream as the first thing designed, not the last.

You will implement Upstream pool, Active health probe, Passive failure counter, Ejection policy, and Pick algorithm, each as a module with one responsibility and a test that proves it in isolation. The traffic distribution pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream.

The walkthrough is written for a developer who has used a traffic distribution 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 upstream pool that returns the right shape, a active health probe that rejects bad input, an integration test that goes green — so you can stop at any point and still have a traffic distribution system that works for the cases covered so far.

What you will build

The finished traffic distribution system exposes five capabilities, each with a focused test:

• Upstream pool
• Active health probe
• Passive failure counter
• Ejection policy
• Pick algorithm

Upstream pool is the foundation that every later stage assumes; Active health probe is where traffic distribution correctness lives; Passive failure counter handles a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream; Ejection policy makes the system observable; and Pick algorithm is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the traffic distribution 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 Upstream pool, Active health probe, and Passive failure counter, with Ejection policy tying them together. Data flows as a list of upstreams each carrying a health state and a rolling failure count, with a picker that skips ejected 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 Passive failure counter does not need to stand up Upstream pool and Active health probe to exercise the interesting branch.

The data model is deliberately explicit. A list of upstreams each carrying a health state and a rolling failure count, with a picker that skips ejected 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 traffic distribution 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 Passive failure counter is designed before the happy path is finished, because a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream is the case that distinguishes a traffic distribution 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 traffic distribution setup is not blocked by a missing tool:

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

The runtime dependencies are chosen to keep the traffic distribution datapath legible:

• fastify (^5.0.0) — Balancer server
• undici (^6.19.0) — Upstream dispatch

# setup for load-balancer-health
pnpm init
pnpm add fastify undici
pnpm add -D typescript tsx vitest

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

Directory structure

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

load_balancer_health/
├── src/index.ts
├── src/upstream-pool.ts
├── src/active-health-probe.ts
├── tests/load_balancer_health.test.ts
└── package.json

Phase 1: Upstream pool

Build the upstream pool that anchors traffic distribution and defines the core data shape.

// Upstream pool: traffic distribution core
export function upstreampool(input: UpstreampoolInput): UpstreampoolResult {
  // Build the upstream pool that anchors traffic distribution and defines the core data shape.
  return processUpstreampool(input);
}

After this phase the upstream pool works in isolation and is covered by a unit test. The upstream pool 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 upstream pool as the single source of truth for the traffic distribution data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Active health probe or Passive failure counter where its effect would be harder to localize.

Phase 2: Active health probe

Add the active health probe so traffic distribution can spreading requests across upstreams correctly.

// Active health probe — handles traffic distribution Add the active health probe
export class Activehealthprobe {
  constructor(private opts: ActivehealthprobeOptions) {}
  async run(ctx: Context): Promise<ActivehealthprobeOutput> {
    return this.execute(ctx);
  }
}

After this phase the active health probe works in isolation and is covered by a unit test. This is where traffic distribution correctness lives: the active health probe is what makes spreading requests across upstreams and ejecting unhealthy ones before clients see failures true, and its test is the proof.

Keep the active health probe 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 active health probe about the transformation.

Phase 3: Passive failure counter

Wire the passive failure counter to handle a health check that without losing data.

// Passive failure counter: traffic distribution stage
export const passivefailurecounter = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the passive failure counter works in isolation and is covered by a unit test. The passive failure counter is the difference between a traffic distribution system that works on the happy path and one that survives a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream.

Design the failure before the success. Decide what passive failure counter does when a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream: reject, retry, shed, or degrade — and write that test first. A traffic distribution system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: Ejection policy

Integrate the ejection policy to make traffic distribution observable and testable end to end.

// Ejection policy for traffic distribution
export async function ejectionpolicy(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

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

Phase 5: Pick algorithm

Finalize the pick algorithm so the traffic distribution system degrades gracefully under overload.

// Pick algorithm: traffic distribution finalization
export function pickalgorithm(state: State): Result {
  validate(state);
  return finalize(state);
}

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

Testing

Each phase has a focused test: upstream pool, active health probe, and passive failure counter are exercised in isolation, then an integration test drives the full traffic distribution pipeline and asserts the a health check that case is handled.

Start with unit tests that pin each component’s contract: the upstream pool returns the documented shape, the active health probe transforms correctly, and the passive failure counter 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 traffic distribution scenario end to end. Assert the happy path first, then deliberately trigger a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A traffic distribution 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 traffic distribution 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 passive failure counter failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the traffic distribution test suite
pnpm vitest run

Deployment and operations

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

The traffic distribution system is a choke point, so it is the right place to reject malformed input before it reaches active health probe or passive failure counter. 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 traffic distribution 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 traffic distribution system, and the lockfile is what makes the audit meaningful.

Observability

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

Watch the passive failure counter failure rate separately from the overall error rate. A jump in passive failure counter failures usually means a health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream, not a bug in active health probe, and the two have different owners and different fixes.

Chart the share of traffic distribution 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 health check that passes against a stub endpoint while real requests fail, hiding a degraded upstream. Reproduce with a minimal fixture, then add a guard test so the traffic distribution system rejects the condition loudly.

Performance regresses under traffic distribution load. Profile with the sampling analyzer, locate the hot path in active health probe, and add a targeted benchmark to lock the regression.

Keep the traffic distribution dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is upstream pool, active health probe, or passive failure counter, instead of guessing from a timeout.

Limitations and next steps

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

The traffic distribution system is single-region by design; extending to multi-region means making upstream pool shareable across regions, which is a real change to the consistency model rather than a config flag. The passive failure counter 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 traffic distribution correctness that is the point.

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