March 3, 2026
Active Incident

AWS Middle East Infrastructure Disruption: Full-Scale Assessment

AWS regions me-south-1 (Bahrain) and me-central-1 (UAE) suffered catastrophic physical infrastructure damage. Three primary BGP prefixes were withdrawn. This report covers the complete scan of all 281,574 hostnames across those prefixes — 30,910 real domains and 250,664 EC2 infrastructure entries — with DNS resolution and HTTP probing for every one.

Regions
me-south-1, me-central-1
ASN
AS16509 (Amazon)
Total Hostnames
281,574
Real Domains
30,910
Report Date
March 3, 2026
30.9K
Real Domains
54%
Operational
46%
Down
8,301
Migrated
91%
Stayed in AWS
Data Composition
281,574 Hostnames: What's in the Dataset
89% of hostnames on the withdrawn prefixes are EC2 reverse DNS entries (bare infrastructure), not customer-facing domains. The report separates these to give an accurate picture.

30,910 Real Domains

Customer-facing websites, APIs, SaaS applications, government portals. These are the domains organizations and users interact with.

54.1% operational

250,664 EC2 Reverse DNS

Entries like ec2-3-29-32-137.me-central-1.compute.amazonaws.com. Bare EC2 instance hostnames — infrastructure plumbing, not user-facing services.

95.1% down

Why This Matters

Reporting "281K domains affected" without separating EC2 infrastructure from real domains overstates the user-facing impact. The 30,910 real domains are the ones that matter to organizations and end users. EC2 reverse DNS entries represent AWS's own infrastructure — important for understanding the compute layer damage, but not the same as customer services going dark.

What Happened
Three BGP Prefixes Withdrawn, Two Regions Dark

Withdrawn Prefixes

PrefixIPsASNRegion
157.175.0.0/1665,536AS16509me-south-1 (Bahrain)
15.185.0.0/1665,536AS16509me-south-1 (Bahrain)
3.28.0.0/15131,072AS16509me-central-1 (UAE)

AS16509 Health During Event

MetricPre-EventCurrentChange
Peer degree117214+83%
Upstream providers16113+606%
Lost upstreams01111 providers dropped
Reputation75/100MULTIPLE_UPSTREAMS_LOST
Current State
30,910 Real Domains: Full DNS + HTTP Scan
Every domain resolved via live DNS and probed via HTTP. Spot-checked with random sampling — 10/10 "down" domains confirmed unreachable.

Domain Status Breakdown

Down — unreachable
11,273
36.5%
Live — restored on same IP
8,136
26.3%
Migrated — live on new IP
6,453
20.9%
Migrated — unreachable
1,848
6.0%
Live — app error (4xx)
1,577
5.1%
Down — NXDOMAIN (gone)
1,072
3.5%
Live — server error (5xx)
551
1.8%

16,717 Operational (54.1%)

Serving HTTP responses: 8,136 restored in-place + 6,453 migrated live + 1,577 app errors + 551 server errors.

14,193 Down (45.9%)

11,273 unreachable on original IPs + 1,848 migrated but unreachable + 1,072 NXDOMAIN (domain gone entirely).

Infrastructure Layer
250,664 EC2 Instances: Compute Damage
EC2 reverse DNS entries represent the raw compute layer. Their recovery rate tells a different story from customer domains.

EC2 Instance Status

226,656 Unreachable (90.4%)
NX
5%

95.1% of EC2 instances remain down. Only 12,341 (4.9%) respond to HTTP. 11,666 are NXDOMAIN — AWS has deregistered those instance hostnames entirely.

The ~5% that respond are likely instances on partially restored sub-ranges that happened to survive or were re-provisioned.

Prefix Recovery
Per-Prefix Restoration Status (Real Domains, Same IP)
Of the ~21,500 real domains still pointing to original AWS ME IPs, what percentage are actually serving HTTP?
PrefixReal DomainsHTTP LiveLive RateStatus
3.28.x.x6,4153,33852.0%Majority Restored
3.29.x.x7,4673,51047.0%Partially Restored
15.185.x.x4,0991,85645.3%Partially Restored
157.175.x.x3,5561,56043.9%Partially Restored

New AWS ME Prefixes (All 100% Live)

PrefixTypeDomains Migrated HereLocation
158.252.x.xNew1,355Dubai Al Barsha
16.24.x.xNew729Zallaq, Bahrain
157.241.x.xNew472Zallaq, Bahrain
16.25.x.xNew300Manama, Bahrain
AWS deployed new infrastructure in-region

2,856 real domains have already migrated to these new prefixes. All are 100% operational. This is AWS's primary recovery strategy — new IP ranges in the same geographic locations (Zallaq, Manama, Dubai Al Barsha).

Migration
Where 8,301 Migrated Domains Went
8,301 real domains changed their DNS to new IPs. The overwhelming majority stayed within AWS and stayed in-region.

Provider Destination

91% of migrated domains stayed within AWS. Only 747 moved to a non-AWS provider.

ProviderDomains%
AWS (all variants)7,55491.0%
Cloudflare2563.1%
Hetzner570.7%
Akamai370.4%
GNX270.3%
DigitalOcean260.3%
Oracle Cloud220.3%
Google Cloud190.2%
Hostinger190.2%
Microsoft Azure160.2%
Alibaba Cloud150.2%
Others (50+)2523.0%

Geographic Destination

Migration overwhelmingly stayed in the Middle East — not the US/Canada pattern seen in smaller samples.

3,740
UAE
Dubai Al Barsha (2,462), Dubai (228)
3,172
Bahrain
Zallaq (1,389), Manama (1,132)
341
United States
Ashburn (63), Seattle (48)
262
Germany
Frankfurt (214)
262
Canada
Toronto (126), Montreal (83)
181
India
Mumbai (170)

+ IE (41), GB (39), SG (36), FI (36), FR (32), ZA (25), HK (22), and 20+ more countries

83% of migrated domains stayed in BH or AE

This is a major correction from the 230-domain sample, which suggested Canada was the #1 destination. At full scale, the migration is overwhelmingly in-region — organizations moved to new AWS infrastructure in the same countries, not to distant regions.

Recovery Strategies
How Real Domains Recovered
StrategyDomains%Mechanism
Restored in-place8,13626.3% Same IP as before, AWS partially re-announced routes and restored infrastructure
New ME infrastructure5,68818.4% Moved to new AWS prefixes: 158.252.x (Dubai), 16.24.x / 157.241.x (Zallaq), 16.25.x (Manama)
AWS other regions1,8666.0% DNS changed to ELBs in other AWS regions (eu-central, us-east, ca-central, ap-south)
App error (4xx)1,5775.1% Server responds but returns client errors — partially recovered or misconfigured
CDN (Cloudflare/Akamai)2930.9% Full migration to CDN anycast. Cloudflare (256), Akamai (37)
Server error (5xx)5511.8% Server responds but returns 500-series errors — infrastructure up, application down
Other providers4541.5% Hetzner, DigitalOcean, Oracle, Google Cloud, Azure, Hostinger, Alibaba, etc.
Unreachable11,27336.5% Still points to original AWS ME IP, HTTP timeout — no recovery action taken
Migrated but down1,8486.0% Changed IP but new destination also unreachable — failed migration
NXDOMAIN1,0723.5% Domain registration expired or DNS removed entirely
Summary
Full-Scale Assessment at a Glance

281K Hostnames, 31K Real Domains

89% of hostnames on the withdrawn prefixes are EC2 reverse DNS — infrastructure, not services. The 30,910 real domains are the customer-facing impact.

54% of Real Domains Operational

16,717 of 30,910 real domains are serving live HTTP traffic — restored in-place (8,136) or migrated to new infrastructure (6,453) or responding with errors (2,128).

46% Still Down

14,193 real domains non-operational. 11,273 unreachable on original IPs. 1,848 migrated but still down. 1,072 NXDOMAIN (domain gone).

91% AWS Lock-In

Of 8,301 migrated domains, 7,554 stayed within AWS. Only 747 (9%) moved to another provider. Cloudflare was #1 alternative at just 256 domains.

83% Stayed in Middle East

Migration was overwhelmingly in-region: UAE (3,740) and Bahrain (3,172). AWS deployed 4 new prefixes in Dubai and Bahrain. Only 17% went outside ME.

Prefix Restoration: 44-52%

For real domains still on original IPs, 44-52% are serving HTTP. 3.28.x leads at 52%. AWS has partially restored all 4 sub-ranges but none is fully back.

Key Takeaway

The raw number — 281K hostnames affected — dramatically overstates the user-facing damage. When you separate EC2 infrastructure (95% down) from real domains (46% down), the picture is severe but not catastrophic. Over half of customer-facing services have recovered, mostly by moving to new AWS infrastructure in the same region. The real risk is the 11,273 domains that haven't taken any recovery action at all — still pointing to unreachable IPs with no DNS change.