On March 1, 2026, a major power outage struck Amazon Web Services’ me‑central‑1 region (Middle East), disrupting EC2 instances, networking APIs, and database services in one Availability Zone (mec1‑az2). The incident stemmed from an unusual physical event where external objects hit a data center, sparking a fire and forcing a full shutdown.
Timeline of Events
- 4:30 AM PST: Outage begins, connectivity issues reported.
- 4:51 AM PST: AWS confirms investigation into power failure.
- 6:09 AM PST: Localized power failure in mec1‑az2 confirmed.
- Afternoon: Engineers deploy configuration changes to mitigate API failures.
- 2:28 PM PST: AllocateAddress API shows recovery.
- 6:01 PM PST: AssociateAddress API restored with forced disassociation capability.
Impact
- EC2 & EBS disruption: Instances and volumes in mec1‑az2 went offline.
- Networking failures: Critical APIs (AllocateAddress, AssociateAddress, DescribeRouteTable) throttled or failed.
- RDS downtime: Databases in the affected zone incapacitated.
- Customer challenges: Elastic IPs trapped in the downed zone couldn’t be reassigned until AWS deployed fixes.
Mitigation & Recovery
- Traffic rerouting: Requests shifted to healthy Availability Zones.
- Forced disassociation: Customers could reassign Elastic IPs to new resources.
- Operational guidance: AWS urged customers to launch workloads in unaffected zones or alternate regions, restoring data from EBS snapshots.
- Provisioning delays: Increased demand in healthy zones led to longer instance launch times.
Lessons for Organizations
- Multi‑AZ architecture is critical: Customers with redundancy across zones were largely insulated.
- Disaster recovery planning: Backup strategies and cross‑region failover are essential.
- Physical risks matter: Even cloud infrastructure can be disrupted by unexpected physical incidents.
- API resilience: Networking dependencies can be single points of failure if not diversified.
Final Thought
The AWS Middle East outage underscores a fundamental truth: cloud resilience depends on architecture, not just provider guarantees. For leaders, the takeaway is clear: design workloads to survive zone‑level failures, and treat disaster recovery as a business‑critical function.
Leave a Reply