AWS says it can't restore some data from mideast facilities struck by Iran
Missile strikes on AWS data centers in Bahrain and the UAE have destroyed multiple availability zones, leaving some customer data permanently unrecoverable despite built‑in regional redundancy. Commenters examine how this outcome collides with AWS’s marketing around “11 nines” of durability, the legal constraints of data residency rules that prevented off‑region backups, and the shared‑responsibility gap where many customers assumed “the cloud” included disaster recovery against war‑scale events. The incident is used as a case study in why organizations still need independent, cross‑region or offline backups, especially in politically unstable regions.
Scope of the Incident & What Was Lost
- Multiple AWS availability zones (AZs) in the me-south-1 (Bahrain) region were destroyed; some AZs in me-central-1 (UAE) also reported down.
- AWS says it “can’t restore some data,” but thread notes it’s unclear which services (S3, EBS, RDS, etc.) are affected.
- There is confusion about whether any multi-AZ S3 data was lost, or only single-AZ resources like EBS and certain S3 classes.
- Status of the Muscat local zone is unclear; region-level health reports show me-south-1 as effectively down.
Durability Guarantees, SLAs, and Force Majeure
- Many point to S3’s “11 nines” durability and multi-AZ design; others highlight footnotes that:
- It’s “designed for,” not guaranteed.
- SLAs exclude force majeure (acts of war, etc.).
- Some argue this makes the headline durability number misleading when war risk dominates the remaining failure probability.
- Counterpoint: you shouldn’t mix “routine” hardware failures with force majeure; they require different mitigation strategies (e.g., full disaster recovery).
Data Residency & Legal Constraints
- Strong data residency rules in Gulf states (and elsewhere) often forbid copying data outside the country.
- AWS says it won’t move data out of chosen regions without customer agreement.
- Some customers wanted to mirror data out after the first AZ hit but were blocked by their own legal/compliance teams.
- EU examples are cited where encrypted offsite backups can be allowed, showing residency isn’t always absolute.
Responsibility for Backups and DR
- Many stress that AWS is a toolbox: customers who didn’t configure cross-region replication or offsite backups accepted the risk.
- Others argue non-technical leadership often assumes “cloud = DR,” underestimating their own responsibilities.
- Classic DR principles (multi-region, multi-provider, offline/cold storage) are repeatedly mentioned but often not implemented due to cost or friction.
Cloud Centralization, War, and Risk Modeling
- Debate over whether centralizing on hyperscalers increases risk by making data centers high‑value wartime targets.
- Some say small self-hosted setups are less likely to be targeted; others counter that geographic diversity, not ownership, is what matters.
- Several argue war should be an explicit failure mode in regions with obvious geopolitical risk.