Build3 publishers2 min readPublished
AWS ran out of options for the data left in Bahrain when the second zone failed
AWS's status update says the strike damage exceeded what its regional and multi-AZ services were built to withstand. Customers who held data in a single zone in Bahrain or in mec1-az2 are being told it is not coming back.
The Engineer · Build desk
What happened
- A status update seen by Reuters says AWS cannot restore access to its Bahrain facility or to one of three data-hosting zones in the United Arab Emirates after damage during the Iran war.
- AWS said at the time that two of its UAE facilities were struck directly, and that a drone strike close to one of its Bahrain facilities physically damaged the infrastructure there.
- After the first Bahrain zone was damaged in March, AWS told customers to migrate to other regions, and most did before a second zone was disrupted in April and the region became unavailable.
- The company said it has assessed all the affected infrastructure and exhausted every option for restoring data and resources that were not migrated before the Bahrain region went offline.
- AWS said it is replacing affected UAE infrastructure and will report on restoration in the coming months, with a further update on Bahrain operations in early 2027.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- constraint By the provider's own account, in-region redundancy does not cover damage on this scale. The remaining customer-side control is a copy held in another region.
- decision Single-zone placement is a durability decision, and teams have to name the volumes and databases that exist in one building only.
- cost Customers absorb the recovery work: AWS said it supported them in re-establishing operations elsewhere, using backups where those existed and alternative solutions where they did not.
The operative phrase in the UAE section of the update is "hosted exclusively in one availability zone" [3]. AWS has not called the UAE region lost; the phrase is about resources, and anything with a single copy in mec1-az2 has nothing left to serve it. Engineers are still working on the wider UAE region and on the other two affected zones [4]. Amazon said most of its UAE customers restored operations in other regions by relying on backups or by copying data that remained accessible [16].
AWS wrote down its own design limit. "The damage to our infrastructure spanned multiple Availability Zones and exceeded what our regional and multi-AZ services are designed to withstand," the company said in the status update, which Reuters was first to report [2]. An availability zone is a cluster of one or more data centres in the same geographic region [21], and it is a subdivision of the network built to provide redundancy [22]. Spreading copies across three zones in one country covers failures that stop at one zone. Satellite imagery showed hits to two Amazon data centres in Bahrain [14].
Bahrain sets the real recovery target: customers had roughly a month between the damage to the first zone and the loss of the region [18]. A region-exit plan that takes a quarter to execute would not have finished.
For "most customers migrated" to describe your own workload, you have to know which of your resources exist in one zone only, at the level of individual volumes and databases. And each of them needs a copy outside the region. Banking operations were among the systems affected when the outages began, Reuters reported, citing a person familiar with the situation [7].
Siting is the part a customer cannot fix. The Islamic Revolutionary Guard Corps said the Bahrain facilities were struck because of Amazon's support for US military operations, according to the account published by the Spokesman-Review [13]. Iran's retaliation was aimed at Israel and at Gulf states hosting US military bases [5]. The UAE is revising plans for a major AI data-centre project, including by considering dispersing facilities and using underground and blast-resistant infrastructure, Reuters reported [12]. Tuesday's status update was the first public one AWS had issued on the damage in five months [15].
What to watch
- Whether the promised UAE update says mec1-az2 has been replaced or written off, and whether the other two zones come back with data intact.
- Whether the authorities AWS notified in Bahrain and the UAE turn this into a rule about where copies of regulated data must sit.
- Whether any customer or insurer puts a figure on the data AWS says it cannot restore.