Massive AWS Outage Disrupts Major Apps; DNS Issue in us-east-1 Blamed
A widespread Amazon Web Services (AWS) outage took major apps and sites offline or made them sluggish, including Alexa, Venmo, Lyft, Snapchat, Canva, Fortnite, Reddit, Disney+, Apple Music, Pinterest, Roblox, banks, airlines, and news outlets. AWS tied the incident to DNS resolution issues impacting DynamoDB API endpoints in the US-EAST-1 (N. Virginia) region.
- Root cause: DNS resolution problems for regional DynamoDB service endpoints.
- Timeline (ET): Errors began overnight (around 3:11 AM); DNS issues were mitigated early morning; lingering impacts affected EC2 instance launches into the afternoon; AWS reported full resolution by evening.
- Knock-on effects: After DNS stabilized, AWS rate-limited new EC2 instance launches in us-east-1 to aid recovery and advised avoiding AZ-specific deployments during restoration.
While most services recovered progressively once the DNS trigger was addressed, cascading dependencies caused elevated errors and connectivity issues across multiple AWS services. One expert likened it to apps being “separated from their data” for several hours—data remained intact, but services couldn’t reliably find it.
Why it matters
AWS accounts for roughly 30% of global cloud infrastructure market share, and US-EAST-1 is a heavily used region. Concentration risk means outages in a single region can ripple across the internet. The incident highlights the importance of multi-region architectures, graceful degradation, and tested failover plans for critical workloads.
Key links and status pages
- AWS Service Health Dashboard
- Reddit Status · Epic Games (Fortnite) Status · Snapchat Status
- Downdetector (service reports)
Takeaways for teams on AWS
- Design for region-level failures (active-active or warm standby across regions).
- Avoid single-AZ coupling and use health checks that can bypass a failing zone.
- Harden DNS and service discovery to tolerate resolver/endpoint issues.
- Run regular game-day failover drills; understand provider rate limits during recoveries.
Even with resilient designs, shared dependencies can cause far-reaching outages. Did this one affect your daily apps or workload operations?
Discussion: What resiliency changes will you prioritize after this outage—multi-region, alternative providers, or improved incident communications?