Virtuozzo Infrastructure 7.3 (7.3.0-171) (formerly Virtuozzo Hybrid Infrastructure)¶
Issue date: 2026-03-31
Applies to: Virtuozzo Infrastructure 7.3
Virtuozzo Advisory ID: VZA-2026-010
1. Overview¶
This release focuses on technical improvements, user experience enhancements, storage performance, and S3 protocol capabilities. It also introduces the product rename from Virtuozzo Hybrid Infrastructure to Virtuozzo Infrastructure. Additionally, this release delivers stability fixes and addresses issues identified in previous releases.
2. New Features¶
[Compute] VM migration management. System administrators can monitor migration progress, cancel or retry migrations, and view the history of migration tasks. For failed migrations, the system shows the exact step where the failure occurred.
[Compute] GPU metering. Added support for creating custom metrics and mapping GPU-enabled flavors to these metrics, enabling GPU usage tracking in GPU-hours.
[Object storage] S3 storage multitenancy. Introduced initial support for aligning S3 storage with the compute tenancy model (project-based isolation). In this release, the functionality is available via the command-line interface. Further enhancements are planned in future releases.
3. Improvements¶
Updated S3 API schemas to support protocol extensions, including Object Tagging.
Improved performance for workloads using small block sizes (for example, 8 KB) with erasure coding, resulting in better storage efficiency and responsiveness.
Introduced a Grafana dashboard that provides improved visibility into the data healing (replication) process, including preliminary estimated completion times (ETAs).
4. Bug Fixes¶
Unsupported baseline CPU models could be offered for a cluster, preventing some VMs from being created. (VSTOR-96428)
License alerts did not respect the automatic license renewal period. (VSTOR-102166)
Deleting a volume snapshot could block the UI until the operation completed. (VSTOR-104120)
Creating a load balancer could fail when Proxy Protocol was enabled together with a UDP listener. (VSTOR-104559)
The CS service could enter a failed state if its repository directory remained locked for too long. (VSTOR-106595)
Upgrade eligibility checks and upgrade tasks could start before update packages were downloaded. (VSTOR-119967)
CSD services with RDMA enabled could fail to recover automatically after a network restart. (VSTOR-120417)
Backup storage throttling could be configured on a node that did not run backup storage services, causing the operation to fail. (VSTOR-120792)
The S3 ACC service could indefinitely retry requests to the name service. (VSTOR-121082)
Stale shaman_node_crash_threshold metrics could remain after a node was released from the cluster. (VSTOR-121148)
A node could remain stuck in the ‘Entering maintenance’ state. (VSTOR-121204)
Removing a compute VLAN network could fail if the VLAN interface was already absent. (VSTOR-121404)
A conflict could occur when resuming a virtual machine that had already been resumed. (VSTOR-121409)
Clients could continue to be redirected to an old upstream for too long after a redirect policy was changed. (VSTOR-121454)
The nova-compute service could get stuck while running nova-qcow2-backup if a required public key was missing. (VSTOR-121902)
More PCI devices than requested could be reserved during VM rescheduling. (VSTOR-122298)
The qemu-img check could fail to repair a corrupted QCOW2 image. (VSTOR-122354)
Object storage alerts did not correctly distinguish between S3 and NFS volumes. (VSTOR-122521)
Downloading updates could fail when /tmp was mounted as tmpfs. (VSTOR-123125)
The system could report an unmount error for a mount point that was already unmounted. (VSTOR-123203)
Users could not dismiss the telemetry consent banner. (VSTOR-123871)
Prometheus 3.3.1 could not collect metrics when Consul was still running an older version. (VSTOR-124089)
The load balancer backend could set an incorrect session persistence type for HTTP cookie-based persistence. (VSTOR-124852)
A race condition could occur in hostd while resolving the current host name. (VSTOR-124904)
VM live resize was not fully restricted in unsupported scenarios. (VSTOR-125086)
Incorrect metadata tier information was displayed in settings after configuring an S3 cluster. (VSTOR-125304)
Incremental backups could fail on heavily loaded clusters. (VSTOR-125509)
The public Barbican endpoint could be used instead of the internal endpoint during load balancer creation. (VSTOR-125563)
A KeyError could occur during live migration after a failed live NIC detach operation. (VSTOR-125622)
The backend could resume Shaman too early during a node update. (VSTOR-125693)
OpenStack Watcher audits could remain pending after a node stopped acting as a controller. (VSTOR-125836)
VM live migration could fail after increasing the number of vCPUs using live resize. (VSTOR-126349)
Node registration settings could be removed even if registration did not complete successfully. (VSTOR-126413)
Flavor lists were not paginated in the Kubernetes worker edit dialog. (VSTOR-126426)
Load balancer ‘Overview’ charts showed incorrect throughput units (traffic displayed in bit/s instead of bytes/s). (VSTOR-126928)