Russia’s Largest Cloud Lost Three Data Centers in Four Days

Yandex confirmed on October 11 that its data center in the Vladimir region is completely shut down after a drone attack. That is the third Yandex facility knocked out inside four days. Sasovo in the Ryazan region went offline on October 8 after a fire. Kaluga lost several modules on October 9.

The Yandex Cloud outage did not stay inside one building. Yandex reported power interruptions in the ru-central1-a availability zone. Yandex Go users could not order taxis in Russia, and the same failures showed up in Serbia, Kazakhstan, Belarus and Armenia. Drivers lost Yandex Pro, internal support systems went unreachable, and smart home users reported outages.

An availability zone is a building, and the buildings share a border

Yandex Cloud publishes five zones across four Yandex data centers. Zone isolation is designed for the ordinary failure, a power feed, a switch, a disk shelf. It assumes the other zone is still standing. Nothing in that model covers a threat that reaches every zone in the same week, with the same weapon, from the same direction.

Put that test against your own estate. Multi-zone and multi-region buy independence only when the failure domains do not share a cause. Sites in three separate regions still share a grid, a regulator, a border and an adversary. The compute is redundant. The geography is not.

The contract moved before the incident review started

On October 10, with two sites already down, Yandex amended its user agreement to list drone and missile attacks among force majeure circumstances. That edit is the honest signal in this story. A provider that priced a physical attack as an outlier would not have written it into the terms of service the day before the third strike.

So read the force majeure clause before the SLA when you buy regional capacity. An SLA credit covers downtime the provider agreed to measure. It does not cover a clause where the provider never agreed to carry the risk at all.

This pattern is not new. When a single provider’s region failed wholesale in Japan, IDCF Cloud told 495 customers to restore from their own backups, because the backups were inside the same region. The same lesson arrived from a different direction in September, when Oracle served force majeure notices over power it could not obtain.

Three questions for your own platform. Which regions share a threat surface that zone separation cannot fix? Which services fail together because they sit on one provider’s footprint? And what does your contract say about war, drones and missiles, in writing, today?

Sources. The Kyiv Independent reported the October 11 strike and the first one on October 8. Yandex published its own account of the Sasovo and Kaluga damage on October 9 and said it could not yet say when that equipment would return.