AWS just admitted the data is gone... #aws #database #programming
Read the Portuguese translation →
Author: Better Stack · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: AWS just admitted the data is gone… #aws #database #programming · Channel: Better Stack · 01:31 · Watch on YouTube
Topics: Data, databases and embeddings, AI infrastructure and hardware
Summary
Section titled “Summary”AWS told some customers that data in Bahrain and one availability zone in the UAE could not be restored after drone strikes damaged data centers. The damage exceeded what AWS’s regional and multi-AZ services were designed to survive. Customers who recovered were able to use backups stored in a different region. The video emphasizes that multi-AZ is an uptime strategy, not a backup strategy, and urges viewers to check where their backups live.
Key points
Section titled “Key points”- AWS said some customer data was permanently gone, not merely unavailable or delayed.
- The transcript says Iranian drones hit AWS data centers in Bahrain and the UAE in early March.
- By the end of April, AWS said restoration could take months.
- On September 15th, AWS said damage in Bahrain spanned multiple availability zones and exceeded what its services were designed to survive.
- The transcript says one of three availability zones in the UAE faced the same problem.
- Availability zones are designed to isolate failures with separate buildings, power, and networking.
- An availability zone is still a data center within the same region.
- Customers who recovered had backups stored in a completely different region.
- Multi-AZ is an uptime strategy, not a backup strategy.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | AWS reports data loss after damage to data centers | The video describes AWS customers being told their data was gone after drone strikes damaged data centers. |
| 00:30 | Damage exceeds multi-AZ protections | AWS said the damage affected multiple availability zones, exceeding what regional and multi-AZ services were designed to survive. |
| 01:00 | Backups must be in another region | The video explains that remote backups helped customers recover and warns that multi-AZ does not replace backups stored in another region. |
Quotes
Section titled “Quotes”Not down, not delayed, permanently gone. — 00:00
Availability zones are supposed to isolate these failures. — 00:30
Multi-AZ is not a backup strategy, it’s an uptime strategy. — 01:00
All you really have then is copies. — 01:31
Concepts and entities
Section titled “Concepts and entities”AWS, availability zone, Multi-AZ, S3
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 AWS just told some of their customers their data was gone. Not down, not delayed, permanently gone. And these workloads were running across multiple availability zones. Let me explain. Back in early March, Iranian drones hit AWS data centers in both Bahrain in the UAE. At first, it just looked like a really bad outage. Then by the end of April, AWS said restoration could take months. But on September 15th, 6 months later, the language changed. AWS said the
00:30 damage in Bahrain spanned multiple availability zones and exceeded what its regional and multi-AZ services were designed to survive. So basically, in other words, anything that only existed in this Bahrain region, it can’t be restored. And one of three availability zones in the UAE also faces the same problem. Availability zones are supposed to isolate these failures. Separate buildings, separate power, separate networking. So if one goes down, your app can keep running in another. And services like S3 advertise this kind of
01:00 durability. But here’s the catch. An availability zone is still a data center in the same region. And some war doesn’t really care about how your architecture diagram actually looks. The customers who recovered were the ones with backups stored in a completely different region. AWS says most of them were able to build elsewhere using those remote backups. Everyone just learned a pretty brutal lesson. Multi-AZ is not a backup strategy, it’s an uptime strategy. And those are two very different things. So if you run anything important in the cloud, check where your backups actually
01:31 live. Because if production and your backups are in the same region, you don’t really have any backups. All you really have then is copies. If you enjoy coding tips and tricks like this, be sure to subscribe to the Better Stack channel.