Virtuozzo Infrastructure 7.4 Hotfix 3 (7.4.0-97)¶
Issue date: 2026-09-28
Applies to: Virtuozzo Infrastructure 7.4
Virtuozzo Advisory ID: VZA-2026-025
1. Overview¶
This update provides stability fixes.
2. Bug Fixes¶
The default gateway over a VLAN could be lost when a bond interface was brought up. (VSTOR-57933)
Restarting the management panel could leave background tasks stuck because their locks were cleared while the task executor kept running. (VSTOR-99864)
Listing network interfaces could take more than 10 seconds on nodes with many VLANs, slowing network pages and tasks in the admin panel. (VSTOR-102226)
Live migration could time out for a VM with a backup NBD backing chain. (VSTOR-103803)
Neutron port checking could consume all CPU resources on a node in clusters with thousands of ports, stalling the OVS agent and forcing a full resynchronization. (VSTOR-106312)
Stability fixes for the object storage service. (VSTOR-109145, VSTOR-139340)
An OVS bridge brought down and back up from the admin panel could remain down, leaving the physical interface detached. (VSTOR-126027)
Stability fixes for the file storage metadata service. (VSTOR-129100, VSTOR-142136)
Rolling compute updates could skip management-only nodes, leaving them on previous container images and the message broker cluster in a mixed-version state. (VSTOR-134183, VSTOR-143632)
Management HA virtual IPs could not be configured when a network had multiple interfaces on a node. (VSTOR-134282)
Unicast traffic on VLAN and flat networks could be flooded to unrelated VMs on the same network. (VSTOR-135180, VSTOR-136328)
The S3 gateway could leak memory when checking multipart uploads. (VSTOR-135296, VSTOR-144832)
Unmounting the storage could hang when a storage node became unreachable, leaving dependent services such as libvirtd stuck. (VSTOR-135626, VSTOR-143108)
A stability fix for the storage command-line tools. (VSTOR-135713)
A VM could hang during shutdown while closing its network backend. (VSTOR-135862)
Kubernetes cluster configuration requests could fail with a gateway timeout during an update. (VSTOR-135949, VSTOR-137444, VSTOR-145141)
The NFS service could abort when two client mounts created a file with the same name at the same time. (VSTOR-136567)
Cluster usage reports could omit CPU cores from storage-only nodes, resulting in incomplete pay-as-you-go usage reporting. (VSTOR-137322)
An HA-protected VM could remain down after evacuation was interrupted during rebuild, while being reported as active with no running domain. (VSTOR-137383)
A recoverable metadata error during background housekeeping could abort all S3 gateways in the cluster, causing a cluster-wide S3 outage. (VSTOR-137637)
Creating a load balancer with a TERMINATED_HTTPS listener could fail during a partial cluster update because the certificate container was rejected. (VSTOR-138621)
Listing compute volumes could fail with an internal error when the service project name was not unique across domains. (VSTOR-139716, VSTOR-144850)
The monitoring auto-updater could overwrite the product version in /etc/hci-release with a newer build than the cluster was actually running. (VSTOR-140456)
The update eligibility check did not flag nodes with an NVIDIA driver older than 595.58.03, the minimum version supported by version 8.0. (VSTOR-141192)
The installer could create an undersized swap partition on nodes with very large amounts of RAM. (VSTOR-141311)
A stability fix for the backup gateway key-value store. (VSTOR-141465)
A kernel stability fix for the storage client after a connection abort. (VSTOR-141615)
After an account was migrated, a gateway node with a warm proxy map and redirection enabled could crash or deny access to the account. (VSTOR-141623, VSTOR-142631, VSTOR-143844)
The product version reported to guests could remain at the previous version after an update until the compute service was restarted. (VSTOR-141752)
Volume image metadata could be missing from backup and volume export snapshots, causing volumes restored to a new VM to lose properties of the original volume. (VSTOR-142022)
Problem reports could be missing cluster-wide data if collection exceeded its time limit or failed on an individual item. (VSTOR-142182, VSTOR-142185, VSTOR-142187)
Problem report archives and JSON files could be read before they were completely written, resulting in truncated files. (VSTOR-142183, VSTOR-143671, VSTOR-143771)
Collected node data could be deleted before being packed into a problem report because the stale-file threshold matched the collection cycle. (VSTOR-142184)
The NVIDIA driver version was shown with a trailing space in the admin panel. (VSTOR-142238)
The backup gateway could miss updates to account information files when detecting changes by modification time. (VSTOR-142291)
A stale storage mount point directory could remain on a node after a software update. (VSTOR-142363)
The backup gateway could not unmount its NFS storage while the storage mount point was being switched. (VSTOR-142375)
A VM with a finalized volume could fail to start after unshelving or evacuation because a stale NBD backing store was written back to its configuration. (VSTOR-142663)
The management API could remain unavailable for several minutes longer than expected while a node rebooted during an update because the database connection pooler did not stop promptly. (VSTOR-143208)
Putting a node into maintenance could fail with an insufficient resources error even if its VMs had already been migrated to other nodes. (VSTOR-143382)
A numeric disk serial number could be replaced with an invalid value in a cluster report, making the report invalid JSON. (VSTOR-143462)
UEFI Windows guests using Microsoft 2023 Secure Boot certificates could fail to boot after migration because the certificates were missing from the firmware variable templates. (VSTOR-143531)
The Windows guest tools VirtIO storage driver could complete an incorrect request under heavy I/O, causing guest crashes and data corruption. (VSTOR-143534)
Backup progress notifications could exceed the message size limit and fail to publish for backups with many chunks. (VSTOR-143590)
Guest I/O could fail after live migration of a VM with an online-resized volume, causing the guest file system to become read-only. (VSTOR-143592)
Creating a storage cluster could take several minutes longer than necessary because an internal task continued waiting for a lock that had already been released. (VSTOR-143595)
The node agent could fail to register after installation. (VSTOR-143694)
The default replication priority boost could overload the network on clusters with fast NVMe drives. (VSTOR-143906)
Proxying between nodes could stop working after the proxy certificate was rotated and remain unavailable until the service was restarted. (VSTOR-144012)
Added a command-line tool for importing an existing StoreOnce datastore into the backup gateway. (VSTOR-144178, VSTOR-144183)
An S3 upload lock that could not be retired could cause the gateway to consume all CPU and eventually run out of memory. (VSTOR-144828)
The S3 gateway upload-lock sweep period could not be configured independently because it used the complete-request sweep setting. (VSTOR-144831)
A stability fix for the storage client mount. (VSTOR-145189)
A false-positive port check warning could appear after the Neutron OVS agent was updated. (VSTOR-145327)
The VM console in the admin panel could show an error after its connection was closed. (VSTOR-145461)
The login and account recovery pages in the admin panel could fail to preload their stylesheets. (VSTOR-145462)
A software update could unnecessarily stop VMs with a virtual TPM because their TPM process was incorrectly treated as outdated. (VSTOR-146060)
The update eligibility check could fail after the monitoring auto-updater installed a telemetry collector newer than the management backend. (VSTOR-146684)