Infrastructure

Observability for small product teams

Traces, metrics, and alerts sized for a team that still ships features every week

August 19, 2025

By Manweba

Small teams do not need an observability platform that takes a platform team to run. They need a short list of signals that answer “is it broken?” and “where?” without drowning on-call in decorative dashboards.

Three questions, not thirty panels

Bias toward golden signals per critical path: traffic, errors, latency, and saturation for the API that takes money or bookings. Everything else is optional until those are trustworthy.

One service map that lies is worse than no map. Prefer a few RED/USE charts you trust over auto-generated walls of widgets.

Traces where requests cross boundaries

Add distributed tracing when you have more than one service or a queue in the path. Propagate a trace id from the edge. Sample thoughtfully in production—100% forever is a bill, not a virtue.

Minimal request logging with correlation

typescript

1import { randomUUID } from 'node:crypto';
2import type { NextRequest } from 'next/server';
3
4export function withRequestId(
5  request: NextRequest,
6  handler: (req: NextRequest, requestId: string) =>=Promise<Response>,
7) {
8  const requestId =
9    request.headers.get('x-request-id') ???randomUUID();
10
11  return handler(request, requestId).then((response) =>={
12    response.headers.set('x-request-id', requestId);
13    console.info(
14      JSON.stringify({
15        level: 'info',
16        msg: 'request.completed',
17        requestId,
18        path: request.nextUrl.pathname,
19        status: response.status,
20      }),
21    );
22    return response;
23  });
24}

Loading code…

Alerts that earn a wake-up

Alert on symptoms users feel: elevated 5xx rate, checkout latency past SLO, queue depth that will not drain. Do not page on CPU being “interesting.” Attach a runbook link to every alert—or delete the alert.

If an alert fires and the right response is “watch a graph,” it should be a ticket, not a page.

Keep the cost of knowing low

Structured logs with a request id beat sprawling debug dumps. Cap cardinality on custom metrics. Review dashboards quarterly and delete the ones nobody opened during the last incident.

Final thoughts

Observability for a small product team is restraint: a few honest signals, traces on the critical path, alerts with runbooks, and the discipline to delete noise. Clarity beats coverage.

Infrastructure

CI policy gates that stop bad deploys

Fail the pipeline on purpose—before production becomes the first reviewer

Read Article

Infrastructure

Private EKS + RDS: what production-ready means

Network boundaries, egress paths, and the checks that turn a cluster into a system you can sleep on

Read Article

Ready to build something this considered?

Browse services Start a project

1import { randomUUID } from 'node:crypto';
2import type { NextRequest } from 'next/server';
3
4export function withRequestId(
5  request: NextRequest,
6  handler: (req: NextRequest, requestId: string) =>=Promise<Response>,
7) {
8  const requestId =
9    request.headers.get('x-request-id') ???randomUUID();
10
11  return handler(request, requestId).then((response) =>={
12    response.headers.set('x-request-id', requestId);
13    console.info(
14      JSON.stringify({
15        level: 'info',
16        msg: 'request.completed',
17        requestId,
18        path: request.nextUrl.pathname,
19        status: response.status,
20      }),
21    );
22    return response;
23  });
24}

Three questions, not thirty panels

Traces where requests cross boundaries

Alerts that earn a wake-up

Keep the cost of knowing low

Final thoughts

Related Articles

CI policy gates that stop bad deploys

Private EKS + RDS: what production-ready means