Overview

The “Overview” screen is the way into a project: a single timeline of what is broken right now and what was recently fixed, gathered from every incident source at once — hosts, uptime monitors, metrics, SLOs, and transaction and profile regressions. Instead of six separate lists, each aware only of its own source, there is one screen answering “how is the project doing”.

Where it lives

The first item in the left rail — “Overview”, URL /projects/<id>/overview. The site root and the project switcher in the header lead here too: pick a project and you land on its Overview. The address /projects/<id>/incident-feed, where the screen lived before version 0.29.0, answers with a permanent redirect (301) to the new one.

Status line

Under the heading are three tiles; each one is clickable as a whole and leads to the matching section.

Three clarifications, without which the numbers are easy to misread:

• Uptime is a single pool: successful and total checks of every monitor in the project are summed and divided, not averaged per monitor. A monitor checked more often weighs more in that figure. Paused monitors are not excluded, and maintenance windows are not subtracted from the denominator — the precision here matches the uptime column in the monitor list, not the single-monitor page, where windows are accounted for. If there were no checks in the window at all, the tile shows “no data”; if the check storage did not answer, it shows “unavailable”, and the rest of the page still renders as usual.
• Hosts over threshold counts distinct hosts, not incidents: three open incidents on one machine make it one. The window switcher does not affect this tile; it is always about what is open right now.
• New issues in 24h counts by the time an issue was first seen. An old issue that got noisy again does not land here — the Issues section, sorted by last event, is for that. The tile’s window is always 24 hours and deliberately ignores the switcher: the number has to stay comparable day over day.

A zero on the second and third tiles means “quiet”. A subsystem you have not set up yet (no monitors, no host agent installed) yields zero or “no data”, not an error — the screen opens either way.

Window: 24 hours or 7 days

The “24h” / “7d” switcher sits by the “Recently resolved” heading, and on an empty screen right below the empty state, so a quiet day can be widened to a week. These are plain links, no JS: “24h” is the screen’s canonical address, “7d” is the same address with ?range=7d. The window defaults to 24 hours; an unknown parameter value is silently read as 24 hours. The choice is stored nowhere and lives only in the address — a link to the weekly window can be bookmarked.

The window affects three things: the “Recently resolved” section, the uptime tile, and the deploy list. It does not affect the “New issues in 24h” and “Hosts over threshold” tiles, or either section of open items — those are about “now” by definition.

Deploys

If the selected window contains deploy markers, a table of the latest ones (up to 20) appears under the status line — version, environment, and when, newest first — with an “All deploys” link to the full section. No deploys in the window means no section at all: not an empty table, an absent block. The reason for the adjacency is simple: most of “what happened” starts with “what we shipped”.

The timeline

Below are three sections. There are deliberately no filters and no pagination on this screen: “Overview” answers “what is going on”, while the dense history of monitor downtime lives on the “Availability incidents” page.

Open groups — up to 50 incident groups, newest first. A group is a root failure (a host went silent, a monitor went down) plus everything that failed because of it; the card expands in place and shows its members. What the badges on members mean and how the grouping works is on the incident groups page.

Ungrouped — up to 50 open incidents that belong to no open group, across all six sources: hosts, uptime monitors, metrics, SLOs, transaction regressions, profile regressions. The root of an open group is not repeated here — it is already shown in its own card’s header. The window does not apply to this section: it is about what is open right now.

Recently resolved — what closed within the selected window: up to 50 groups and up to 50 standalone incidents. Those are two independent caps, and the caption under the heading says so plainly — the section can show up to a hundred cards and rows combined.

Rows are laid out the same way in every section: source, name, status, start time, and resolution time.

When the screen is empty

If all four queries come back empty at once — no open groups, no open ungrouped incidents, and nothing resolved within the window — the sections are replaced by a “Getting started” prompt leading to the project setup page. The status line, deploys, and the window switcher stay where they are: an empty timeline means “nothing had a chance to break”, not “the screen is broken”.

Access

The screen is available to any member of the organization the project belongs to — there is no separate permission for it. A project in someone else’s organization answers 404, not 403: the status code must not reveal whether such a project exists.