Routing & Internet
How the internet runs: BGP and RPKI, route leaks and hijacks, DNS, peering, subsea cables, outages and their postmortems.
Recent activity
Stories per weekLatest stories
6 most recent-
Philippines subsea cable cut rerouted within hours without full outage
October 2 subsea cable cut between Philippines and Singapore caused intermittent connectivity and reduced international bandwidth for multiple telcos. PLDT and Converge rerouted traffic through alternate cables within hours; Globe fully restored service by October 5. Cable segment handled ~300Gbps of traffic; rerouting capacity proved sufficient to avoid complete outage.
Why it matters Real-world case study of subsea cable failure recovery; demonstrates multi-path redundancy and rapid failover limit blast radius and enable recovery without prolonged service loss.
-
Azure gateway outages hit 18 regions in 40 hours
Between September 29 and October 1, 2026, Azure experienced two separate outages that knocked gateway services offline across 18 regions within 40 hours, affecting Azure OpenAI customers in Sweden for nearly six hours; root cause involved infrastructure servicing but Microsoft delayed postmortem publication.
Why it matters Back-to-back regional gateway failures expose timing-of-interaction risks in multi-region deployments; lack of detailed postmortem limits ability to prepare defensive routing and failover strategies.
-
Subsea cable fault slows internet across the Philippines
The Department of Information and Communications Technology is investigating a major subsea fiber-optic cable outage on October 2, 2026, affecting internet connectivity across the Philippines; authorities are evaluating technical degradation, environmental factors, and potential sabotage.
Why it matters Recent subsea cable incidents (Red Sea September 2026, Philippines October 2026, West Africa 2024) expose single-point-of-failure risk in regional architectures; understanding failure patterns directly impacts BGP rerouting strategy and peering decisions for Asia-Pacific operators.
-
RPKI is not a significant source of BGP noise
APNIC analysed 11 years of RPKI and BGP data to measure routing activity associated with RPKI-related changes. Study finds RPKI is not currently a significant source of BGP noise despite growing ROA adoption and ROA management activity.
Why it matters Operators deploying Route Origin Validation can proceed with confidence that RPKI deployment will not destabilize the routing system at current adoption rates.
-
Race condition with schema migration broke Railway routing globally
Railway's global routing service returned 404 errors for five minutes after a race condition caused a new routing service to deploy before its required database schema migration completed. Service queried non-existent columns, causing all route lookups to fail across all regions.
Why it matters Illustrates critical deployment ordering and validation challenges for distributed routing control planes; demonstrates why schema validation must precede service rollout in mission-critical infrastructure.
-
Azure ExpressRoute and VPN Gateway fail across 19 regions
Microsoft Azure experienced a major network outage on September 30 affecting 19 regions, impairing ExpressRoute and VPN Gateway connectivity. Root cause was a recent regional gateway management service change coinciding with unrelated infrastructure OS maintenance.
Why it matters Demonstrates cascading failure risk when control-plane changes align with maintenance windows; enterprises must segregate management and data-plane maintenance schedules.