Cloudflare logged six separate incidents on 18 November 2026, touching Access, network performance in Paris, a multi-datacentre disruption across India, R2 downloads in Japan, GraphQL API latency and a near five-hour connectivity problem between Europe and Asia Pacific. All six were Cloudflare-side faults, not anything you could have fixed on your own infrastructure.
Started
01:10 UTC
Duration
Lasted 1h 31m
Source
IsDown
Next time
Cloudflare is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: Cloudflare Sites and Services, R2, D1, Cache Reserve, Asia, Hyderabad, India - (HYD), Ahmedabad, India - (AMD), Patna, India - (PAT), New Delhi, India - (DEL), Mumbai, India - (BOM), Nagpur, India - (NAG), Kanpur, India - (KNU), Cloud Network Interconnect (CNI), Network, Paris, France - (CDG), Europe, Access
What happened?
Cloudflare filed six incidents across the day, starting at 01:10 UTC and running until the last one closed just after 00:36 UTC the following day, a span of nearly 24 hours. The first hit Access, where Cloudflare said customers might see delays making configuration changes, resolved in 1h 31m. A second, shorter incident at 03:39 UTC named Network Performance in Paris, FR, though Cloudflare's own closing update narrowed the impact to Cloudflare Network Interconnect (CNI) only, lasting 1h 14m. At 08:04 UTC Cloudflare reported a network disruption across multiple datacentres in India, naming Hyderabad, Ahmedabad, Patna, New Delhi and Mumbai among the affected locations, resolved in 45 minutes. A separate R2 issue in Japan caused downloads to stall for users connecting via Japanese datacentres, running 1h 19m with only two updates, investigating and then monitoring, and no explanation of root cause in either. Later, Cloudflare identified latency on GraphQL Analytics API queries at 20:35 UTC and marked it monitoring five minutes later, a 2h 50m incident by the record despite the quick update cadence. The longest incident of the day began at 22:26 UTC: degraded connectivity between Europe and Asia Pacific, affecting R2, D1 and Cache Reserve along with anything else routed on that path, lasting 4h 55m before Cloudflare confirmed a fix. Across all six incidents Cloudflare published 17 updates total, and none of them stated a cause beyond "the issue has been identified," which tells you Cloudflare found something internally but chose not to publish what. No user reports were filed during the window on this page. The component list for the day spans 17 entries, from named cities to whole continents, which shows how widely a single day's problems can be spread across Cloudflare's network.
Learning
UptimeRobot's own probes against Cloudflare's api-control, dns-1111 and edge-data components, run from EU and NA regions every 60 seconds, show 100% uptime over 24 hours, 7 days and 30 days, and this incident was not registered against them. That is not a contradiction. Several of the day's problems were geographically or functionally narrow, a stalling R2 download path in Japan, a network interconnect fault scoped to Paris, degraded routing specifically on the Europe-to-Asia-Pacific path, and a check running from EU and NA endpoints against different components can sail through green while those specific paths are broken. This is the standard shape of a partial network fault: the provider's edge or interconnect layer fails for a defined route or region, and a synthetic check sitting outside that path has no visibility into it at all. Watching your own application endpoints tells you your service responded, it does not tell you whether the request took a degraded path through Paris or across the Pacific to get there. Cloudflare's 90-day history shows 211 incidents with a median duration of 1 hour 56 minutes, so a day with six incidents and one running nearly five hours is not unusual for this provider. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.