---
title: Components and groups
description: Choose which monitors your users see, and organise them into groups that make sense to them.
sidebar:
  label: Components & groups
---

{/* Source of truth: packages/shared/src/status-pages.ts —
    ComponentStatuses, componentGroups (name, monitorIds, sortOrder). */}

A **component** is a monitor exposed publicly under a name your users
understand. A **group** collects related components under a heading.

## Naming

The component name is the public label. Name it for what the user does, not for
what you run.

| Internal monitor | Component name |
| --- | --- |
| `prod-api-lb-health` | API |
| `pg-primary-tcp` | Database |
| `auth0-login-flow` | Sign-in |
| `s3-upload-canary` | File uploads |

If a user cannot tell whether a component's outage affects them, the name is
wrong.

## Groups

Each group has a name, a set of monitors, and a sort order. Typical shapes:

- **By product area** — Web app, API, Integrations
- **By region** — Europe, North America, Asia-Pacific
- **By audience** — Customer-facing, Developer platform

Groups are worth adding once you have more than about six components. Below
that they add ceremony without helping.

## Ordering

Put the components most users care about first. The overall banner tells people
whether *something* is wrong; the top of the list should tell them whether it is
the thing they use.

## What to leave off

Every component you add is a public statement about your architecture, and one
more thing to explain when it goes red.

Leave off internal-only dependencies, staging, anything with a monitor that
flaps, and anything whose failure a customer would never notice. Those still
belong in Glowo — just on the dashboard, not the status page.

## Overriding status during an incident

Component status normally follows the monitor. During an
[incident](/status-pages/incidents) you can set it explicitly, because the human
picture is often more accurate than the probe's.

A monitor that checks one health endpoint may stay green while a feature behind
it is broken for everyone. Setting `PARTIAL_OUTAGE` says what your users are
actually experiencing.