Stop 11 · Monitoring & ObservabilityTopics · 2/9

Metrics, Logs & Traces (The 3 Pillars)

In short

When something breaks, you'll ask three things: how bad is it, what exactly happened, and where did the time go? Metrics, logs and traces each answer one of those, and you need all three.

Metrics are numeric measurements over time, logs are discrete event records, and traces follow a single request across multiple services — together they give a complete picture of system behavior.

The mascot, as a doctor, holds a heartbeat chart and a notepad beside footprints leading through hospital rooms.

Think of it like this

A patient in hospital

The BP chart shows numbers over time. The nurse's notes record each event: 'fever at 2 am, gave medicine'. The patient's path shows each department visited and how long each took. Metrics are the chart, logs are the notes, traces are the path.

Watch it happen

One failed checkout, three signals

checkout svc

Metrics

Logs

Traces

errors_total +1

1/5A counter goes up. Cheap to store, great for graphs and alerts, but no details about this one request.

The big ideas

What you need to know

Metrics: numbers over time

Counters, gauges and histograms. Cheap at scale, ideal for alerts and trends, light on detail.

Logs: events with detail

What happened, with context. Use structured JSON so you can search by field, not by regex.

Traces: one request's path

Spans show each hop and its timing across services. Best for 'where did the time go?'

Link them with IDs

Put the trace ID in every log line. Then you can jump between signals during an incident.

See it in code

Walk through the code

json

1{2  "time": "2026-10-27T10:42:13Z",3  "level": "error",4  "msg": "payment failed",5  "order_id": "ord_8812",6  "trace_id": "4bf92f3577b34da6a3ce929d0e0e4736",7  "duration_ms": 12408}

1/3When, how serious, and a short fixed message you can search and group by.

A structured log line. Most logging libraries for Node.js, Python and Go can write JSON like this, and OpenTelemetry can add the trace ID for you.

Save yourself the pain

Common mistakes

Oops

Putting user IDs or order IDs as metric labels.

Do this instead

Each unique label value creates a new time series and can overload your metrics system. Put IDs in logs and traces.

Oops

Logging free-text sentences like 'oops it broke for user 7'.

Do this instead

Log structured fields with a fixed message, so you can filter by level, user or order in seconds.

Your mission

Instrument one endpoint

~30 min

Quick check

You need to know which of five services made one request slow. Which signal answers that best?

Go deeper

1Identify a real debugging scenario and which pillar (metrics, logs, or traces) would help most.
2Learn why traces are especially critical in microservice architectures where one request touches many services.
3Understand these three pillars are complementary, not redundant with each other.