阅读中文版本

Build an Alerting Rule Evaluator

Alerting Rule Evaluator — a complete walkthrough of alert evaluation, covering rule loader, metric reader, condition evaluator, and testing.

alerting_rule_evaluator/
├── src/index.ts
├── src/rule-loader.ts
├── src/metric-reader.ts
├── tests/alerting_rule_evaluator.test.ts
└── package.json

// Rule loader: alert evaluation core
export function ruleloader(input: RuleloaderInput): RuleloaderResult {
  // Build the rule loader that anchors alert evaluation and defines the core data shape.
  return processRuleloader(input);
}

// Metric reader — handles alert evaluation Add the metric reader so
export class Metricreader {
  constructor(private opts: MetricreaderOptions) {}
  async run(ctx: Context): Promise<MetricreaderOutput> {
    return this.execute(ctx);
  }
}

// Condition evaluator: alert evaluation stage
export const conditionevaluator = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// Cooldown state for alert evaluation
export async function cooldownstate(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Notification dispatcher: alert evaluation finalization
export function notificationdispatcher(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a alert evaluation system whose reason for existing is evaluating threshold and ratio rules over metric streams and firing alerts without flapping. The temptation in alert evaluation 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 rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter as the first thing designed, not the last.

You will implement Rule loader, Metric reader, Condition evaluator, Cooldown state, and Notification dispatcher, each as a module with one responsibility and a test that proves it in isolation. The alert evaluation pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter.

The walkthrough is written for a developer who has used a alert evaluation 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 rule loader that returns the right shape, a metric reader that rejects bad input, an integration test that goes green — so you can stop at any point and still have a alert evaluation system that works for the cases covered so far.

What you will build

The finished alert evaluation system exposes five capabilities, each with a focused test:

• Rule loader
• Metric reader
• Condition evaluator
• Cooldown state
• Notification dispatcher

Rule loader is the foundation that every later stage assumes; Metric reader is where alert evaluation correctness lives; Condition evaluator handles a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter; Cooldown state makes the system observable; and Notification dispatcher is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the alert evaluation 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 Rule loader, Metric reader, and Condition evaluator, with Cooldown state tying them together. Data flows as a set of rules each with a condition over metric queries and a cooldown window that suppresses re-firing.

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 Condition evaluator does not need to stand up Rule loader and Metric reader to exercise the interesting branch.

The data model is deliberately explicit. A set of rules each with a condition over metric queries and a cooldown window that suppresses re-firing 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 alert evaluation 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 Condition evaluator is designed before the happy path is finished, because a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter is the case that distinguishes a alert evaluation 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 alert evaluation setup is not blocked by a missing tool:

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

The runtime dependencies are chosen to keep the alert evaluation datapath legible:

• fastify (^5.0.0) — Rule API
• undici (^6.19.0) — Metric query
• redis (^4.7.0) — Cooldown state

# setup for alerting-rule-evaluator
pnpm init
pnpm add fastify undici redis
pnpm add -D typescript tsx vitest

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

Directory structure

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

alerting_rule_evaluator/
├── src/index.ts
├── src/rule-loader.ts
├── src/metric-reader.ts
├── tests/alerting_rule_evaluator.test.ts
└── package.json

Phase 1: Rule loader

Build the rule loader that anchors alert evaluation and defines the core data shape.

// Rule loader: alert evaluation core
export function ruleloader(input: RuleloaderInput): RuleloaderResult {
  // Build the rule loader that anchors alert evaluation and defines the core data shape.
  return processRuleloader(input);
}

After this phase the rule loader works in isolation and is covered by a unit test. The rule loader 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 rule loader as the single source of truth for the alert evaluation data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Metric reader or Condition evaluator where its effect would be harder to localize.

Phase 2: Metric reader

Add the metric reader so alert evaluation can evaluating threshold and ratio correctly.

// Metric reader — handles alert evaluation Add the metric reader so
export class Metricreader {
  constructor(private opts: MetricreaderOptions) {}
  async run(ctx: Context): Promise<MetricreaderOutput> {
    return this.execute(ctx);
  }
}

After this phase the metric reader works in isolation and is covered by a unit test. This is where alert evaluation correctness lives: the metric reader is what makes evaluating threshold and ratio rules over metric streams and firing alerts without flapping true, and its test is the proof.

Keep the metric reader 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 metric reader about the transformation.

Phase 3: Condition evaluator

Wire the condition evaluator to handle a rule that flaps without losing data.

// Condition evaluator: alert evaluation stage
export const conditionevaluator = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the condition evaluator works in isolation and is covered by a unit test. The condition evaluator is the difference between a alert evaluation system that works on the happy path and one that survives a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter.

Design the failure before the success. Decide what condition evaluator does when a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter: reject, retry, shed, or degrade — and write that test first. A alert evaluation system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: Cooldown state

Integrate the cooldown state to make alert evaluation observable and testable end to end.

// Cooldown state for alert evaluation
export async function cooldownstate(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

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

Phase 5: Notification dispatcher

Finalize the notification dispatcher so the alert evaluation system degrades gracefully under overload.

// Notification dispatcher: alert evaluation finalization
export function notificationdispatcher(state: State): Result {
  validate(state);
  return finalize(state);
}

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

Testing

Each phase has a focused test: rule loader, metric reader, and condition evaluator are exercised in isolation, then an integration test drives the full alert evaluation pipeline and asserts the a rule that flaps case is handled.

Start with unit tests that pin each component’s contract: the rule loader returns the documented shape, the metric reader transforms correctly, and the condition evaluator 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 alert evaluation scenario end to end. Assert the happy path first, then deliberately trigger a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A alert evaluation 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 alert evaluation 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 condition evaluator failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the alert evaluation test suite
pnpm vitest run

Deployment and operations

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

The alert evaluation system is a choke point, so it is the right place to reject malformed input before it reaches metric reader or condition evaluator. 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 alert evaluation 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 alert evaluation system, and the lockfile is what makes the audit meaningful.

Observability

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

Watch the condition evaluator failure rate separately from the overall error rate. A jump in condition evaluator failures usually means a rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter, not a bug in metric reader, and the two have different owners and different fixes.

Chart the share of alert evaluation 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 rule that flaps around its threshold and fires repeatedly because the cooldown is shorter than the metric jitter. Reproduce with a minimal fixture, then add a guard test so the alert evaluation system rejects the condition loudly.

Performance regresses under alert evaluation load. Profile with the sampling analyzer, locate the hot path in metric reader, and add a targeted benchmark to lock the regression.

Keep the alert evaluation dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is rule loader, metric reader, or condition evaluator, instead of guessing from a timeout.

Limitations and next steps

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

The alert evaluation system is single-region by design; extending to multi-region means making rule loader shareable across regions, which is a real change to the consistency model rather than a config flag. The condition evaluator 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 alert evaluation correctness that is the point.

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