## Virtuozzo Infrastructure 7.4

### Overview

This release updates managed Kubernetes to version 1.35.3 and Fedora CoreOS to version 43, introduces API credentials for external automation workflows, and provides the updater patches and tooling required to upgrade the product to version 8.0. It also includes object storage performance improvements, virtual router reliability enhancements, and migration service improvements.

### What's new

#### Compute service

- Kubernetes updated to version 1.35.3. The Fedora CoreOS image used for managed Kubernetes clusters has also been updated to version 43.20260413. This returns managed Kubernetes to a version that is actively supported by the upstream project.

#### Security

- Users can now create API credentials (commonly known as API tokens) for external automation tools such as Terraform, Ansible, and CI/CD pipelines. API credentials are available to every user and can be managed in both the admin and self-service panels.

#### Updates

- Upgrade path to version 8.0. This release includes the updater patches and configuration updates required to perform an in-place upgrade to version 8.0. Hardware support policy has been clarified in relation to the RHEL 10 upstream rebase in version 8.0. A hardware compatibility checker tool is now available on GitHub to verify that your cluster hardware is supported before initiating the upgrade.

### Improvements

- Improved propagation of virtual router and floating IP configuration changes. Virtual router state is now restored more reliably after a cluster outage.
- Optimized performance when adding VMware sources containing hundreds of VMs and improved VM listing efficiency. Added support for migrating VMs located within VMware folders and nested subfolders.
- Certain alerts now take node maintenance mode into account and will not trigger unnecessarily.

### Fixed issues

- [VSTOR-109558] The admin panel could not update nodes added at an older patch level after the cluster was upgraded.
- [VSTOR-120570] The storage agent could hang when accessing failed multipath devices, making nodes unmanageable.
- [VSTOR-128013] Neutron L3 agent router processing could take too long in environments with many routers, causing routing outages.
- [VSTOR-128885] HA destruction could be allowed on mixed-version clusters, leaving compute services in an inconsistent state.
- [VSTOR-129518] The admin panel could remain slow or unresponsive when the backend was overloaded without restarting automatically.
- [VSTOR-130654] Cluster update could fail while moving VMs to the storage mount point if a volume backup was still running for a VM disk.
- [VSTOR-131130] Prometheus could run on all nodes after an update instead of only on management HA nodes.
- [VSTOR-131185] A stability fix for the Intel ice network driver.
- [VSTOR-131859] Self-service UI operations could hang when telemetry endpoints were unreachable.
- [VSTOR-132221] Volume attachment updates could fail after a backup job failed on a removed compute node.
- [VSTOR-133743] Public network ports could be blocked by loop protection during node reboot.
- [VSTOR-134252] Revert to snapshot could be unavailable in the self-service panel for volumes on external storage.
- [VSTOR-134259] Load balancer metrics collection could become extremely slow due to database bloat.
- [VSTOR-134410] VM migration requests could return HTTP 500 when a migration was already in progress.
- [VSTOR-134774] Validating a migration source cloud could time out when credential storage was accessed through an external address that was not available on the cluster network.
- [VSTOR-134950] Security fixes for CVE-2026-35385, CVE-2026-35386, CVE-2026-35387, CVE-2026-35388, and CVE-2026-35414 in Kubernetes cluster VM templates.
- [VSTOR-135357] Floating IP connectivity could stop working, making VMs and load balancers unreachable from outside the cluster.
- [VSTOR-136219] Logging in through an identity provider (IdP) to another domain could display the wrong domain while keeping the previous session.