On June 21st, between the hours of 06:27 UTC and 07:42 UTC, Cloudflare experienced a major service disruption across 19 of its datacenters. During this time, depending on your location, you may have been unable to access Hasura services (Dashboard, APIs, Console, etc.) including your project’s GraphQL API.
In summary, a configuration change being deployed by Cloudflare engineers to their Multi-Colo PoP (MCP) locations, resulted in a re-ordering of terms which caused Cloudflare to withdraw a critical subset of prefixes (Where prefixes are a collection of adjacent IP addresses which are advertised to peers on the global network). Due to this withdrawal, our service which is routed via Cloudflare’s network, unfortunately experienced a period of inaccessibility for a large portion of our user base.
Even though the 19 datacenters which experienced the issue account for 4% of Cloudflare’s total network, these datacenters handle a total of about 50% of the total traffic that flows via CF’s backbone.
A more detailed blog has been published by Cloudflare, which details the issue and the sequence of events. Please see the blog here: https://blog.cloudflare.com/cloudflare-outage-on-june-21-2022/