Cloud Platform: еmergency maintenance

We have completed emergency maintenance activities aimed at updating the Cloud Platform infrastructure components.

The activities were performed following the identification of a critical vulnerability in the Linux kernel (CVE-2026-53359) affecting the KVM component used for virtualization.

Successful exploitation of this vulnerability could potentially compromise isolation between virtual environments. In a successful exploitation scenario, an attacker could perform a virtual machine escape and gain access to resources of the virtualization host, creating a potential security risk to data and other virtual machines running on the same infrastructure. Therefore, we promptly initiated remediation activities to mitigate the identified risk.

As part of the remediation process, we updated the affected infrastructure components. Applying the updates required virtual machine reboots to activate the changes. Following completion of the technical activities, the vulnerability was remediated.
The maintenance activities were performed on July 10 and 11 in a phased manner, with continuous monitoring of the impact on customer service availability. The maximum expected downtime for an individual virtual machine was up to 5 minutes. No indicators of compromise affecting customer environments were identified.

Additional information on the use of Nested Virtualization functionality
Following the update of the Cloud Platform infrastructure components, the use of Nested Virtualization functionality is no longer available.
If the use of Nested Virtualization is critical for your business workloads, we recommend considering migrating your applications and services to dedicated servers or Managed Kubernetes clusters running on dedicated servers.

Thank you for your understanding. We apologize for any inconvenience caused.

Emergency maintenance will be carried out on the Cloud Platform infrastructure throughout the business day. Virtual machines will be rebooted as part of the work, with up to 5 minutes of downtime.

More details are available on the status page.