VirtConsole 3.3.4 makes keeping a fleet current a single action, keeps long operations running through service restarts, shows each migration step as it happens, keeps your panels live, and improves NVIDIA GPU support.
Update to latest
An administrator can now update the Master and then every node from one action on the Hypervisors page. Nodes are updated one at a time, you can follow progress and cancel, and a node that still has work in flight is left alone. Each hypervisor page also has its own Update to latest control.
Long operations that stay on course
Image and ISO downloads, image creation, deploys, backups, database operations and migration transfers keep running through service restarts and show progress while they run. Migrations show each step as it happens and tell an operator which step is running. Cold migration supports AlmaLinux and Rocky Linux nodes, and the admin API can start a migration without a storage mapping: shared storage stays in place and local disks move to the storage you name.
VPCs on demand
Create a VPC with or without a NAT gateway. Its router is built on each node as soon as a member instance lands there, so nodes without members carry nothing.
Live panels
Admin and user pages follow changes over websockets for many more resources, including backups, disks, images, VPC, DNS, volumes and scaling groups. The hypervisor page follows storage, GPU and instance changes and shows statistics for each node.
Improved NVIDIA GPU support
GPU plans get smarter about the node they run on. VirtConsole now recognises NVIDIA GPUs that are busy with workloads on the node itself and leaves them alone, and when an instance is destroyed its GPU goes back to the host driver it had before, where that driver is known. The create pages show the GPU a plan requires and whether one is free in the location, and the plan APIs and the OpenTofu plan data sources carry the same information. The hypervisor GPU tab shows each GPU’s memory and links to the instance that holds it. GPU plans are not available for Kubernetes clusters yet.
On KVM nodes, changing an instance’s plan can also add, change or remove its GPU. An added GPU is reserved right away and attached at the next start. Removing or changing a GPU is done on a stopped instance, and the GPU stays reserved until the node confirms it is released.
Source address filter for public interfaces
A new setting in Settings turns on a source address filter for public interfaces. It is off by default. Each instance has a Source check switch for its VPC interface and one for its public interface, on the instance Networking tab and through the API, so workloads such as floating IPs or routers can be exempted. Changes reach the node within about a minute, without a restart.
Networking and first boot
- Windows instances on KVM nodes apply Remote Desktop and the first-logon setup on first boot and keep a stable IPv6 link-local address.
- Subnets are checked against the Linux bridges each node reports, and new deployments go only to nodes that have the subnet’s bridge.
- Deploy requests tell you why a location cannot take them, with distinct status codes for a node that refuses and a node that is unreachable.
- The VPC create page lists every VPC-ready location.
Nodes and Kubernetes
- Node enrollment keeps the node and shows a warning with the reason when the final step cannot reach it, so you can retry from its page. Each node also reports its clock, and a clock out of step with the Master shows as a warning.
- Kubernetes load balancers can reuse certificates from your account with
ssl-mode=certificate. The load balancer certificates tab lists every certificate the load balancer uses, and the certificates page shows the certificate ID in its own column.
Search and filters that search everything
Many search boxes and filters in both panels now run on the server over the whole result set.
Also in this release
- First-boot network configuration covers instances with several public interfaces and IPv6 allocations, with one default route per address family.
- Backups to a mounted backup storage are written directly into it, so full backups of large disks need no temporary space on the node.
- System backup policies can run on a named time zone, summer and winter time included.
- ISOs show which nodes already hold them, and creating an instance while its image is still downloading on the node returns a clear retry hint.
- Long hostnames are truncated with a tooltip on instance tables, and RAM and size displays are rounded to two decimals.
Thanks for building on VirtConsole. Questions along the way, reach us any time.