Amazon Web Services (AWS) has been unable to restore access to some of the data in its infrastructure in the Middle East that was affected by the conflict. This involves the AWS region in Bahrain and one Availability Zone in the United Arab Emirates.
The problems began in March, when AWS data centers in both countries were hit by drone attacks during the conflict with Iran. This resulted in physical damage to buildings, power supplies, and other infrastructure. In some cases, firefighting efforts also caused water damage. AWS warned at the time that a full recovery could take a long time.
More than half a year later, it appears that part of the infrastructure cannot be restored. In Bahrain, the damage affected multiple Availability Zones. According to AWS, the extent of the damage exceeded the design capabilities of its regional services and facilities, which are distributed across multiple zones. Following an investigation, the company determined that it can no longer restore access to resources and data that were stored exclusively in the Bahrain region.
AWS had already advised customers to migrate workloads to other regions following the initial outage in March. Many customers did so before a second disruption in April rendered the Bahrain region completely unavailable. AWS continues to provide support to affected customers as they migrate to other regions.
Different situation in the UAE
In the United Arab Emirates, the situation is less severe. AWS operates three Availability Zones within the me-central-1 region there. The company has determined that resources and data hosted exclusively in mec1-az2 can no longer be made accessible. For regional resources and data in the two other affected zones, mec1-az1 and mec1-az3, recovery efforts are still underway.
That distinction is somewhat obscured in Bloomberg’s reporting, which states that Amazon can no longer access data that was stored exclusively in data centers in Bahrain and parts of its network in the UAE. AWS’s own status reports indicate more specifically that in Bahrain, the entire region was affected, while in the UAE, the inability to restore data is currently limited to a single Availability Zone.
Most customers in the UAE have since resumed their operations elsewhere. To do so, they were able to restore backups or copy data that had remained accessible despite the outage.
Limits of redundancy
At the same time, these events highlight the limitations of the Availability Zone model. AWS divides its infrastructure within a region into separate zones so that an outage at a single location does not automatically affect the entire region. In Bahrain, however, multiple zones were damaged. AWS states that the damage there exceeded what its regional and multi-AZ services were designed to handle.
This means the cause differs fundamentally from a regular technical outage. The infrastructure was physically damaged at multiple locations due to military violence. AWS is working to replace affected equipment and infrastructure and expects to provide more clarity in the coming months regarding further recovery in the UAE. A new update for Bahrain is scheduled for early next year.