System status

• No failed checks
• Some failed or slow checks
• Major outage (under 95%)
• No data

No incidents recorded yet.

• A scheduled job on Cloudflare's network checks each component every 2 minutes (GPU supply every 10). A failed check is retried once, 2 seconds later, before it counts as down. A working answer slower than 2 seconds counts as degraded, and degraded checks count as up in the uptime figures.
• The site and console also run on Cloudflare, so these checks share their network. A Cloudflare-wide outage could stop both the site and its checks, and this page wouldn't show it. A separate check from outside Cloudflare (GitHub Actions, every 15 minutes) alerts us independently.
• For the website and console, response times are measured inside Cloudflare, so they leave out the trip from your network.
• Days are UTC. Grey days mean no data, either before tracking began or when no checks were recorded. We never fill gaps in as working.
• Components marked incident log only aren't checked automatically. Their status changes only when we post an incident.
• These are measurements, not a service-level agreement.

Incident feed (Atom) for a feed reader or Slack's RSS app, and /status.json for scripts: current status and uptime per component, refreshed about once a minute. See Support for how to reach us.