Rogers outage: 17 hours of BGP instability
A major outage at Canadian ISP Rogers Communications started July 8, 2022, around 08:45 UTC and lasted over 17 hours, cutting off a large portion of Canadian internet traffic. Cloudflare Radar data shows a near complete loss of traffic from Rogers' autonomous system (AS812), with only residual traffic observed during the incident.
Based on the telemetry patterns and similar past incidents, this appears to be an internal network error rather than a cyber attack.
BGP withdrawals point to internal failure
BGP is the routing protocol that lets networks advertise their presence to each other. When a network stops announcing its prefixes, other networks simply cannot find it. Rogers was not advertising its prefixes, so its network effectively vanished from the internet.
Cloudflare's data shows a clear spike in BGP update messages after 08:15 UTC, reaching a peak at 08:45 — exactly when the prefix withdrawals from Rogers began.
A second chart showing prefixes received from Rogers in Toronto confirms the withdrawal at 08:45, with a slow recovery attempt starting at 14:30. That attempt was short-lived, with another round of withdrawals around 15:45.
Core network flapping
By 23:00 UTC, the outage had lasted 15 hours with no full recovery. There was a brief bump in traffic around 18:15 — about 3% of usual levels — lasting 30 minutes before dropping again.
BGP announcement data shows Rogers repeatedly attempting to re-advertise its prefixes, with peaks at various intervals, but each attempt was followed by withdrawals moments later. The latest attempt at that time was at 21:45 UTC.
The pattern suggests that internal sessions in Rogers' core network were flapping — causing withdrawals when sessions dropped and advertisements when they came back up. Rogers' Senior Vice President Kye Prigg had confirmed that the root cause had not yet been identified, and there was no estimate for restoration.
Slow recovery begins
By July 9 at 01:50 UTC, partial recovery was visible. Traffic from Rogers started increasing after 00:15 UTC, reaching about 13% of the usual rate at 00:50 UTC, then climbing to roughly 17.5% at 01:50. Frequent BGP flapping continued, indicating the core network issue had not been fully resolved.
Recovery gained momentum through the morning. At 08:40 UTC on July 9, traffic was around 76% of the same time the previous day, and Rogers services were approaching normal levels.
Outages of this scale, while disruptive, are not unprecedented. BGP failures have taken down major networks before — Facebook suffered a multi-hour outage in October 2021 that similarly saw its routing announcements disappear. Cloudflare's Q2 2022 internet disruption summary provides more context on how frequently these incidents occur across the internet's interconnected structure.
For real-time monitoring of internet disruptions, Cloudflare Radar offers ongoing visibility into AS-level traffic and BGP activity. Live updates on incidents are also posted on @CloudflareRadar on Twitter.
