Between October 8 and 11, 2026, drones hit three Yandex data centers in Russia. The outage took down more than 80 services, from banking apps to companies' cloud databases. The lesson applies to businesses everywhere: if your data lives with only one provider, you don't have a backup. You have a single place where you can lose it.
We have grown used to thinking of the cloud as indestructible. In reality, it is buildings full of servers. They have a street address, a power supply and cooling. Like any building, they can be damaged by fire, flood, power failure or attack.
Over four days, drones struck three Yandex data centers. The attacks are attributed to Ukraine, although Ukraine has not officially claimed them.
| Date | Location | Impact |
|---|---|---|
| Oct 11, 2026 | Vladimir | Data center (40–50 MW) shut down; Yandex Cloud (Compute, Object Storage, Kubernetes, PostgreSQL) and more than 80 services went offline (Kyiv Post) |
| Oct 9, 2026 | Kaluga | “Several data center modules are completely disabled” (The Moscow Times) |
| Oct 8, 2026 | Sasovo (Ryazan region) | Fire, facility fully offline; Yandex has not confirmed whether the equipment can be restored (CNBC) |
The outage did not only affect Yandex's own services. Banking apps, food delivery, ride-hailing and the retail chains Magnit and Fix Price, all built on Yandex infrastructure, stopped working too. As late as October 29, three weeks later, Yandex said it had restored only “key features”. It also withdrew its 2026 financial guidance (The Moscow Times).
A cloud provider is responsible for keeping the service running. Making sure your data survives is your responsibility. This split is called the shared responsibility model, and AWS, Azure, Google Cloud and smaller providers all use it.
A cloud with a single provider does not protect you from these risks:
The Yandex case combines two of these: physical destruction and geopolitics. Companies that ran everything on Yandex Cloud alone waited weeks for their services to come back.
The proven foundation is the 3-2-1 rule, now extended to 3-2-1-1-0:
| Number | Rule | Protects against |
|---|---|---|
| 3 | Three copies of the data (production + two backups) | Failure of a single copy |
| 2 | Two different storage types or providers | Failure or outage of one technology or company |
| 1 | One copy at a different location (offsite) | Fire, flood, destruction of a data center |
| 1 | One copy offline or immutable | Ransomware, deletion by an attacker or by mistake |
| 0 | Zero errors when verifying restores | A backup that cannot be restored in a crisis |
In practice, this means spreading backups across three levels:
For critical systems, a multi-cloud or hybrid cloud architecture is also worth it. In an emergency, the application can then be started at another provider, not just have its data restored.
A backup is only half the job. The other half is being able to restore data quickly when the main service goes down. Without that, a backup can turn out to be incomplete, outdated or unreadable in a crisis.
Every system needs two numbers:
A disaster recovery plan should also cover the scenario where an entire provider goes down, not just one server. It should state:
The plan must be tested regularly. At least once a quarter, do a live restore of a selected system into a standby environment and measure whether you met the RTO. Only a test shows whether the backup really works.
The Yandex outage is not just a Russian problem. A data center can be knocked out by a flood, a fire, a network failure or a provider's mistake anywhere in Europe. The cloud is a great tool, but it does not take responsibility for your data's safety off your hands.
A company with backups at different providers, in different locations and with a tested recovery procedure can survive even the outage of an entire provider. Others may wait weeks, like Yandex's customers in October 2026.