Yandex data center in Kaluga goes down: some services may be unavailable

On 9 October 2026, Yandex warned users that some of its services may be unavailable because of an incident at its data center in Kaluga. The company's press office told Russian Business that emergency services are working at the site. Yandex did not state the cause, the scale of the impact, or when service will be restored.

In its headline, RB.RU calls the Kaluga facility "another" company data center that has suspended operations. The report does not explain what happened in the earlier episode. For that reason, it is not yet possible to compare the incidents or look for a link between them: there is no data to do so.

What is known so far

There are few facts, and it is worth keeping them strictly separate.

• Source: Yandex's press office, quoted via Russian Business
• Location: the company's data center in Kaluga
• Status: operations suspended
• On site: emergency services
• Impact: some services may be temporarily unavailable; which ones has not been disclosed
• Not disclosed: the cause, whether anyone was hurt, an estimate of when service will be restored

The mention of emergency services points to a physical incident rather than a software failure. The report does not make clear what kind. It could have been a fire, a power problem or something else. Without confirmation from the company, any version is guesswork, and it is best not to draw conclusions from it.

How a physical incident differs from a software failure

A software bug is usually rolled back: engineers restore the previous version, restart nodes or reroute traffic. A physical incident works differently. While emergency services are working at the facility, engineers' access to the server halls is restricted. Power or cooling may be shut off deliberately to avoid worse consequences. After that come hardware checks and a phased restart, which take time even when the hardware is undamaged.

Large operators build these risks into their architecture. They spread services across several sites and shift traffic to the others when one fails. But redundancy does not work the same way for every component. Frontends and stateless services move quickly. Storage and databases are harder: they need replication between sites, which can be synchronous or asynchronous, and that determines whether users are affected. So Yandex's wording that "some services may be unavailable" is entirely expected. Some products will ride out the loss of a site unnoticed; others will not.

What to watch next

The next few hours will show three things.

• The list of affected services. If the outage hits only certain products, the rest have been successfully moved to other sites.
• The cause of the incident. It will determine whether this is a one-off or a reason to revisit physical security requirements for facilities.
• How long the site takes to return to service. If equipment has to be inspected or replaced, the timeline could be days rather than hours.

Yandex's own analysis is also worth waiting for. For a large infrastructure company, a public postmortem with a timeline and conclusions usually matters more than the first statements.

Practical steps for users and businesses

Ordinary users can simply wait: retrying logins and restarting apps does little if the problem is on the site's side. Companies whose business processes depend on Yandex services, including its cloud platform, should do several things.

• Check the company's status pages and official channels rather than relying on rumors on social media
• Find out which availability zones your resources are in, and whether you have a copy in another zone
• Do not launch mass restarts or aggressive request retries: they add load during recovery
• Make sure current backups are stored outside the affected site
• Prepare a message for customers in advance in case the outage drags on

Any data center incident is a reminder that a service's reliability depends not only on its provider but also on how the customer has built their own architecture. Multi-zone setups cost money, but they are usually cheaper than downtime.