Status pages

A public page for one project: components fed by uptime checks and cron monitors, 90 days of bars, incidents, maintenance and email updates for subscribers.

A status page tells your customers whether your product is working, without signing in to anything. It lists the parts of your product they care about, such as Website and API, with each one's status and 90 days of uptime bars, plus the incidents you post and the maintenance you plan. Visitors can subscribe by email, and scripts can read the same page as JSON or a feed.

Each project has at most one status page, and none until you create one. A page for several products is a project of its own, with its own checks.

Create a page

Open Status page in the dashboard menu, next to Uptime and Crons, and pick a slug. Editors and owners create and change the page; everyone who can see the project, and read keys, can read it. Or from the terminal, with the Little Friend CLI:

A status page from the CLI
littlefriend status create acme
littlefriend status show

Every status page command is under littlefriend status. The page lives at https://status.littlefriend.io/<slug>.

  • Slug. 3 to 50 lowercase letters, digits and dashes, starting and ending with a letter or digit, with no two dashes in a row. It is unique across Little Friend, and a few words such as api, status and www are reserved.
  • Name, up to 80 characters, the project's name by default, and an optional description of up to 300 characters under it.
  • Website link, optional: where the name and logo link to.
  • Logo, optional: PNG, JPEG or WebP, up to 40 KB. SVG is refused. Little Friend serves it from the page's own address, so a visitor's browser never loads an image from anywhere else.
  • Published, off by default. Until you publish it, the page answers not found to everyone but the project's editors and owners, who see it with a preview banner while signed in to the dashboard.

Components and sections

A component is one thing visitors see a status for, such as Website, API or Checkout. Group components in sections, up to 20 per page, or leave them above the sections. A page has up to 50 components.

Each component can follow up to 10 sources from the same project: uptime checks and cron monitors. A component with no sources is manual: it is operational unless an incident, maintenance or an override says otherwise. Deleting a check or monitor removes it from every component.

How a component gets its status

A component is one of these:

  • operational: Operational
  • degraded: Degraded performance
  • partial_outage: Partial outage
  • major_outage: Major outage
  • maintenance: Under maintenance

The first rule that applies decides it:

  1. Override. An editor can set a component's status by hand. It holds until it is cleared.
  2. Maintenance. A maintenance window in progress that covers the component makes it maintenance.
  3. Sources and incidents. The worst of what its sources say and the impact of each open incident that names it.

An uptime check counts as degraded after one failed run and down after 2 in a row. A cron monitor counts as down while its latest runs were missed, failed or timed out. A paused check or monitor, or one that has not run yet, counts for nothing. From its sources alone, a component is a major outage when every source that counts is down, a partial outage when some are, degraded when one is degraded, and operational otherwise.

The page's status is the worst of its components, and any problem outranks maintenance. Its banner reads All systems operational, Degraded performance, Some systems are down, Major outage or Maintenance in progress.

Uptime bars

Each component shows the last 90 days as one bar per day on the project's clock, today last. A narrow screen shows the last 30. A day's color comes from the share of its checks' runs that passed:

BarRuns that passed that day
Operational99.5% or more
Degraded performance98% to under 99.5%
Partial outage90% to under 98%
Major outageunder 90%

One failed run of a 5 minute check in a day is 99.65%, still operational. A day with a missed, failed or timed out cron run is at least degraded, and since check-ins are kept 30 days, a monitor's bars go back 30 days. A day an incident naming the component was open is at least that incident's impact. A day with no runs, no check-ins and no incident is grey. Runs during maintenance count like any other.

Beside the bars, the component's uptime is the share of its checks' runs that passed over the 90 days. It is blank for a component without checks or runs.

Incidents

Post an incident when something is wrong. It has a title of up to 120 characters, the components it affects, and an impact that each of them takes while it is open:

  • degraded: Degraded performance
  • partial_outage: Partial outage, the default
  • major_outage: Major outage

An incident comes with its first update, and you add more as you learn. Each update is plain text up to 5,000 characters, up to 100 per incident, shown newest first. Each update sets the incident's status, and may change its impact and components:

  • investigating: Investigating
  • identified: Identified
  • monitoring: Monitoring
  • resolved: Resolved

A resolved update records when the incident ended. A later update that is not resolved opens it again.

Every update emails the page's confirmed subscribers, once the page is published. Turn off notify on an update to post it without email.

The page shows open incidents at the top, and incidents resolved in the last 14 days under Past incidents. Its history lists 90 days, and each incident has its own page with every update.

Automatic incidents

Automatic incidents are off by default. Turn them on, and an uptime check that fails 2 runs in a row opens an incident for each component it is behind, such as "Website is not responding", as investigating. Its impact is a major outage when every check of the component is down, and a partial outage otherwise. Its update says the automated checks found the component not responding, and never names the check, its URL or its error.

None opens for a component that has an override, is in maintenance, already has an open automatic incident, or had one resolve in the last 30 minutes. That quiet time keeps a check that keeps coming and going from opening an incident every few minutes.

Once every check of the component has passed 2 runs in a row, the incident gets an update saying the component responds again. It is resolved when nobody posted to it; if someone did, it keeps the status they gave it. Both updates are emailed like any other. Cron monitors never open an automatic incident.

Maintenance

Schedule a maintenance window with a title, an optional note, the components it covers, and a start and end. A window is at most 14 days long and starts at most 365 days ahead. It is scheduled before it starts, in progress while it runs and completed after, unless you cancel it.

The page shows windows in progress at the top and the rest under Scheduled maintenance. While a window runs, its components show maintenance, and the uptime checks behind them open no check down alert. A failed run during the window is still recorded and counts toward the streak, so a check still failing when the window ends alerts on its next failed run. An alert already open carries on as usual, and cron alerts are not held. Maintenance sends no email.

Subscribers

Visitors subscribe with their email address in the page's form, which works without JavaScript. The answer is the same whether an address is new, waiting or already subscribed, so the form never tells anyone who is on the list.

  • Double opt-in. A new address gets one email with a confirmation link that works for 7 days, and nothing else until it confirms. The link opens a page with a Confirm button and never confirms on its own, so a mail scanner that opens links cannot subscribe anyone. An address waiting to confirm gets a new link at most every 10 minutes.
  • Emails. Each update with notify on sends one email to every confirmed subscriber, with the incident's title and status, the update, the affected components and a link to the incident. A subscriber gets at most 10 an hour, and the rest wait. An email not sent within 24 hours of its update is dropped.
  • One-click unsubscribe. Every email has an unsubscribe link, and mail clients that offer their own unsubscribe button use it in one click. Unsubscribing deletes the address.
  • Limits. Up to 2,000 confirmed subscribers per page; addresses nobody confirms never use them up. 10 subscribe requests an hour from one visitor's address, and 5 confirmation emails an hour to one address and 200 per page.

Editors and owners see subscriber addresses in full and can remove one. Viewers, read keys and AI assistants see them masked, such as f•••@example.com.

JSON and feeds

A page serves these under its address, all without sign-in and cached for up to 30 seconds:

  • /: the page
  • /history: incidents of the last 90 days
  • /incidents/<id>: one incident with every update
  • /index.json: the page as JSON, readable from any origin
  • /feed.atom: the Atom feed
  • /feed.rss: the RSS feed

The JSON holds the page's status, its sections and components with their status, uptime and daily bars, open and recent incidents with their updates, and maintenance. It answers any origin, so another site or status tool can read it:

Read a page as JSON
curl https://status.littlefriend.io/acme/index.json

The Atom and RSS feeds carry the latest 50 incident updates and maintenance windows, for feed readers and chat tools that follow a feed.

Custom domain

A page can also live on your own host name, such as status.example.com. Enter it on the Status page screen, and Little Friend gives it a token. Then add two DNS records at your DNS provider:

  • A CNAME from your domain to status.littlefriend.io, which routes visitors to the page.
  • A TXT record named _littlefriend-status.status.example.com holding littlefriend-status=<token>, which proves the domain is yours.
DNS records
status.example.com.  CNAME  status.littlefriend.io.
_littlefriend-status.status.example.com.  TXT  "littlefriend-status=<token>"

Press Verify. It looks up both records and says which it found. The TXT record is what verifies the domain, and a domain verified on one page cannot be verified on another. Changing the domain starts over with a new token. Once verified, every path on the domain is the page: status.example.com/history is its history, and nothing else is reachable there.

For now, the Little Friend team sets up the HTTPS certificate for a custom domain after it is verified. Once Verify shows the domain verified, email hello@littlefriend.io with the domain, and we will tell you when it serves over HTTPS. The page at https://status.littlefriend.io/<slug> keeps working in the meantime.

Privacy

The public page shows only the words you wrote: the page, section and component names, incidents, updates and maintenance, with statuses and uptime shares. It never shows a check's name or URL, a monitor's name or slug, an error, a status code or a run's message, and nothing about your visitors.

The page itself sets no cookies and runs no scripts. It is plain HTML with one stylesheet, in light or dark by the visitor's setting, and it loads nothing from anywhere else. It sends no referrer when a visitor follows a link, and other sites cannot frame it. When a project is deleted, its page is gone from that moment.