पाठ 1 / 35

Monitoring vs Observability

यह पाठ अभी हिंदी में अनुवादित नहीं है, इसलिए अंग्रेज़ी में दिखाया जा रहा है।

Explain what observability means, how it differs from classic monitoring, and why distributed systems need it.

लेखक: · 15 साल से वेब डेवलपर

What problem are we solving?

When your code runs on your laptop, you debug it with print statements and a debugger. In production you cannot pause a live server or attach a debugger to a request that failed ten minutes ago. You only know what the system wrote down while it ran. That written-down evidence is called telemetry.

Monitoring is watching a fixed set of things you decided in advance: is CPU above 90%, is the disk full, is the health check green. It answers known questions. Observability is a property of the system: how well you can answer new questions about its internal state using only the telemetry it emits, without shipping new code. A system is observable when an engineer can go from "checkout is slow for some users" to "requests from one region hit a cold cache on one database replica" by slicing existing data.

Monitoring is still needed; observability is what lets monitoring alerts be debugged quickly. The difference matters most in microservices, where one user click touches many processes.

The telemetry journey

Every observability setup, small or large, follows the same path. Code is instrumented to emit signals, an agent or SDK collects them, a backend stores them, engineers query them on dashboards, and rules alert a human when something breaks.

Five connected stages in order: instrument the code, collect telemetry, store it in a backend, query and visualise it, alert a human.
Figure 1.1: instrument -> collect -> store -> query -> alert

The hospital patient chart

A doctor cannot open a patient up every hour. Instead nurses record temperature, pulse and notes on a chart at regular intervals. Monitoring is the alarm that rings when pulse crosses a threshold. Observability is how rich that chart is: with good records, a doctor arriving at 3 a.m. can work out why the patient worsened without re-running every test.

Observability is not a tool you buy

Students often say "we added Grafana, so we are observable". A dashboard tool only displays data. If the service never records request IDs, durations or error causes, no tool can recover them later. Observability starts in your code: what you log, measure and trace.

त्वरित जाँच: Which statement best describes observability?

  • A dashboard that shows CPU and memory for every server
  • The ability to answer new questions about a system's internal state from the telemetry it already emits
  • A paging system that wakes engineers when a threshold is crossed
  • Running a debugger against production processes
Answer

The ability to answer new questions about a system's internal state from the telemetry it already emits — Observability is about answering questions you did not anticipate, using existing telemetry. Dashboards and paging are parts of monitoring that build on it.