/status is the place to look when something on the site seems
broken. It has three parts, and they come from three different places — which
is the thing worth knowing, because they can disagree for good reasons.
Incidents and maintenance
The top of the page is written by people: incidents and maintenance windows staff have published, each with an impact level, the parts of the site it touches, and a timeline of updates. The overall banner is the worst impact among what is open:
| Level | Means |
|---|---|
| Operational | Nothing is currently reported as broken. |
| Maintenance | Planned work; some things may be briefly unavailable. |
| Minor | Something is degraded; most of the site is unaffected. |
| Major | Something is badly degraded and being worked on. |
| Critical | A part of the site is unusable. |
Each event has its own link, so it can be shared while it is running.
The network monitor (Defcon)
The middle of the page is measured. Defcon is a small fleet of monitoring nodes, run from different networks and locations, that load the site’s pages, API routes and hosts the way a visitor would and report what they saw:
- a status per monitor —
OK,DEGRADED,DOWNorUNKNOWN— with a dot for each node that measures it; - uptime over 24 hours, 7 days and 30 days, and the 24-hour average latency;
- a latency chart per monitor with one line per node, because the point of a fleet is the same route seen from everywhere. A browser monitor charts page load time, with its paint milestones (FCP, LCP) and stability (blocking time and layout shift) under it;
- for a traceroute monitor, the latest route hop by hop.
One slow request does not turn a monitor red
A status only gets worse after several consecutive bad samples and only recovers after several good ones, per node. That is deliberate — a single slow response from one place is noise — but it means a problem you are seeing right now can take a few minutes to show here, and a fix a few minutes to clear it.
Green here does not mean it works for you
Defcon measures from its nodes. A problem between your network and the site
— your ISP, a regional route, a filter — is invisible to it. If /status is
green and the site is not loading for you, the traceroute and per-node charts
are the place to look for where the path differs.
Staff decide how much of this is public: the whole section can be switched off, a monitor can be kept private, and the node list, open incidents and traceroutes each have their own switch. Nothing public names a node’s address or a monitor’s full URL.
The same data is available from a terminal with
tmc defcon — including tmc defcon status --check,
which answers with an exit code a script or another monitor can use.
Health checks
The bottom of the page is what the site says about itself, measured automatically and refreshed about once a minute: how much of the game-server fleet is being scanned as often as its policy asks, how far behind the scan queue is, servers shown online that have missed several scans, how far behind analytics are, failed scheduled jobs and new errors. These are the checks behind “why is this server’s player count stale?” more often than an incident is.