Third-party service status

The public status page of a provider you depend on (Cloudflare, GitHub, Atlassian, AWS…), read as a monitor. It fails when the provider reports an incident, so you hear about it before your users do, and a status page can show the provider next to your own services as an external component.

Two formats are read:

• the JSON of an Atlassian Statuspage, /api/v2/summary.json: most public status pages (Cloudflare, GitHub, Atlassian, Discord, DigitalOcean…);
• an RSS or Atom incident feed: the AWS Health feed (https://status.aws.amazon.com/rss/all.rss, or one feed per service and region) and the history.rss of a Statuspage.

Microsoft 365 is not covered: its service health is not public. It is only available through the Microsoft Graph API (serviceAnnouncement) with an application registered in your tenant. Any public RSS or Atom feed of incidents can be given as the address instead.

Network ports to open

From the DumbMonit server (or its relay agent) to the device:

Port	Protocol	Purpose
443	TCP	Status page of the provider (HTTPS)

What it watches

Besides the common availability metrics (probe_success, probe_duration_seconds, probe_failure_info), all prefixed dumbmonit_:

Metric	What
probe_provider_impact	0 operational, 1 minor, 2 major, 3 critical
probe_provider_incidents	open (unresolved) incidents
probe_provider_components_affected	components that are not operational (Statuspage only)

The monitor fails (probe_success at 0, reason status) when the impact reaches Fail on: any degradation by default. A planned maintenance is not an outage. The built-in “service down” alert therefore applies to a provider like to any other service.

With a Component filter, only the components whose name contains the text count (case-insensitive): “Workers”, “Git Operations”, “EC2”. A filter that matches nothing is shown as a configuration error, not as an outage.

In a feed, only the newest entry within Feed window (24 hours by default) counts: while it is not marked “resolved” or “operating normally”, an incident is open, and a “disruption” or “outage” in its text counts as major. Narrow a feed that mixes services with Component.

If the status page itself cannot be read, the last answer is used for up to 15 minutes; after that the monitor fails with the reason connect, timeout, status or body.

The answer is shared: several monitors, or several status pages, on the same provider cost one request per Refresh interval (60 seconds by default). The address cannot point at the DumbMonit host itself unless “Allow loopback and link-local targets” is ticked.

On a status page

A provider monitor added to a status page appears as a normal service tagged “External”, with its own uptime history.

Watch a provider’s status page

1. Enter a provider name (cloudflare, github, atlassian, aws, discord, npm…; the full list is in the guide), or the address of a Statuspage page or RSS/Atom feed.

2. To watch one part of the provider only, enter its name under Component, for example “Workers” or “EC2”.

3. Choose what makes the monitor fail: any degradation, a partial outage or worse, or a major outage only.

Warning

The monitor reads what the provider chooses to publish, often minutes after the fact: it complements your own checks and never replaces them.

No credential is needed: these pages are public.

Credentials

None.

Names: cloudflare, github, atlassian, discord, dropbox, npm, digitalocean, twilio, zoom, vercel, netlify, datadog and aws (AWS Health feed).

Address: a name from the list, the address of a Statuspage page (https://www.githubstatus.com) or of a feed. Options: Component, Fail on (minor, major, critical), Refresh interval (60 s, from 30 to 3600), Feed window (24 h, from 1 to 168) and the usual timeout (5 s).