构建结构化日志管道

项目概述

本项目实现一个面向 log ingestion 的完整系统，核心要解决的问题是 accepting structured logs from many sources, enriching them, and routing them to sinks without blocking writers。整体由 Ingest buffer、Enrichment stage、Route table、Sink writers、Backpressure shed 五个模块组成，每个模块单一职责、可独立测试，最后通过集成测试验证端到端正确性。设计上把 a slow sink that fills the buffer and forces the pipeline to drop logs from the most chatty source 作为第一类需求先行处理，而非事后再补。

架构与关键决策

数据模型为 a queue of log events flowing through enrichment and routing stages, with each sink written asynchronously and shed under load。Ingest buffer 定义其余阶段共同依赖的数据形状并在边界处校验输入；Enrichment stage 承载 log ingestion 的正确性逻辑；Route table 专门处理 a slow sink that fills the buffer and forces the pipeline to drop logs from the most chatty source，在失败路径上给出明确的拒绝、重试或降级策略；Sink writers 提供可观测性，每次操作输出带关联 id 的结构化事件；Backpressure shed 在过载时按既定方式降级，避免过载演变为数据损坏。配置与代码分离，使风险变更可经 diff 审阅。

实施与测试

按五个阶段依次实现 Ingest buffer、Enrichment stage、Route table、Sink writers、Backpressure shed，每阶段配套聚焦单测。集成测试组合各模块并驱动真实 log ingestion 场景：先验证正常路径，再刻意触发 a slow sink that fills the buffer and forces the pipeline to drop logs from the most chatty source 并断言系统降级而非损坏。所有依赖均在进程内或被桩替换，用例在 CI 中确定可复现。

限制与后续

• 单区域设计，多区域复制超出范围。
• Ingest buffer 默认进程内，未水平共享。
• 无鉴权层，假定部署在强制身份的边缘之后。

代码示例

# setup for log-pipeline-structured
pnpm init
pnpm add fastify pino undici
pnpm add -D typescript tsx vitest

log_pipeline_structured/
├── src/index.ts
├── src/ingest-buffer.ts
├── src/enrichment-stage.ts
├── tests/log_pipeline_structured.test.ts
└── package.json

// Ingest buffer: log ingestion core
export function ingestbuffer(input: IngestbufferInput): IngestbufferResult {
  // Build the ingest buffer that anchors log ingestion and defines the core data shape.
  return processIngestbuffer(input);
}

// Enrichment stage — handles log ingestion Add the enrichment stage so
export class Enrichmentstage {
  constructor(private opts: EnrichmentstageOptions) {}
  async run(ctx: Context): Promise<EnrichmentstageOutput> {
    return this.execute(ctx);
  }
}

// Route table: log ingestion stage
export const routetable = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// Sink writers for log ingestion
export async function sinkwriters(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Backpressure shed: log ingestion finalization
export function backpressureshed(state: State): Result {
  validate(state);
  return finalize(state);
}

# run the log ingestion test suite
pnpm vitest run