ProxCenter v1.4.9 Release

ProxCenter v1.4.9 is a feature release focused on safer rolling updates, broader migration sources, high availability and day-to-day operations. It introduces warm migration from XCP-ng, rolling updates you can watch and approve node by node, PBS 4 compatibility, and the public API reference rendered inside the product.

---

🔄 Rolling updates

Watch and approve every node

Rolling updates now show live apt progress per node, with the apt output displayed on screen when a step fails. Each node can require an explicit manual approval, from the wizard or from the Task Center, and a cancel is honoured while the run is paused.

Finished runs stay in the Task Center with their full log, and a cancelled run is reported as cancelled.

Smarter evacuation and checks

• The parallel migrations setting is honoured, and evacuated guests are spread by projected load and affinity rules
• The Ceph wait only holds on real recovery, not on ProxCenter's own noout flag or on unrelated warnings
• Nodes are always reached on their management address, never on a Corosync or Ceph network
• Repository preflight reads the Proxmox fields correctly, and a 2 GB disk floor replaces a free space check that was far too strict
• Every wizard parameter gets a tooltip, alongside health tiles and a warning about NOPASSWD: ALL sudoers rules

---

🚚 Migration

Warm migration from XCP-ng

ProxCenter can now connect directly to an XCP-ng pool through XAPI, alongside Xen Orchestra, and run warm migrations from XCP-ng using CBT over NBD.

The former XCP-ng Live mode has been removed, and offline downloads now survive a slow SSH poll and retry up to three times.

Hyper-V, Nutanix and Windows guests

• Hyper-V inventory is collected in a single PowerShell call, wrong credentials are reported as an authentication error, and disk paths on SMB shares are resolved
• Hyper-V and Nutanix guests keep their source CPU count and memory instead of defaulting to 1 vCPU and 2 GB
• Compact OS Windows guests convert: the preflight installs the ntfs-3g system-compression plugin

Less disk space, clearer errors

• Cold migrations to file storage write the disk once: the converted image is adopted in place, cutting temporary space from 2x to 1x
• Warm node preparation states that it requires Proxmox VE 9 instead of returning an apt transcript
• Installing missing packages gets a 20 minute budget, and a 401 on the enterprise repository no longer aborts Debian packages

---

🖥️ Consoles

VNC and SPICE consoles gain fullscreen and scaling controls, remembered per console type. The SPICE console now offers VM power actions (start, shutdown, stop, pause) with automatic reconnect, and noVNC adds a send-key menu and clipboard text.

Keystrokes typed into a guest are also no longer logged to the browser console.

---

🛡️ High availability and DRS

One leader at a time

On an HA control plane, the orchestrator lock is now taken on the PostgreSQL primary only, and the connection pool refuses a hot standby. A single orchestrator acts as leader at any time.

DRS from any node

The DRS page is populated from whichever node serves it: metrics and recommendations are routed to the leader through a new /api/v1/leader probe.

• A long HA preflight returns its verdict instead of a false "Orchestrator unavailable"
• DRS migrations stranded in a running state after an orchestrator restart are settled automatically
• The task footer shows the real percentage instead of a fixed 50%

---

🛟 Site Recovery

• A replication job continues past a failing VM and lands in a new Partially synced status, with the failed VMs named
• A replicated VM carrying a Proxmox snapshot syncs again instead of failing on rbd: File exists
• Alerts and reports understand the partial status, and the failure alert no longer flaps during a rerun

---

💾 PBS 4 compatibility

ProxCenter is now compatible with Proxmox Backup Server 4.x: package refresh, repository toggle, API token listing, S3 endpoints and prune jobs all work again.

---

📖 The API reference inside ProxCenter

The public API reference is now rendered inside the product, under Settings, with Try it requests sent to your own instance. Exploring and testing the API no longer requires leaving ProxCenter.

---

✨ Also in this release

Area	What's new
🌐 Network	sFlow configuration succeeds, reports per-node results and is re-applied after a node reboot; flows attributed by guest MAC on SDN topologies
🔔 Notifications	Commas accepted in the default recipients field, and lists saved with the semicolon workaround repaired
🎨 Interface	Tooltips stay reachable on disabled buttons; the storage overview table fills the page
📦 Dependencies	Twenty-five Dependabot updates, including undici 8, cron-parser 5, dagre 3 and react-resizable 4

---

⬆️ Upgrading

One database migration is included and runs automatically at first boot. It adds the XCP-ng connection sub-type, and existing connections are backfilled as Xen Orchestra.

docker compose pull && docker compose up -d

HA clusters

DRS leader routing needs a new HAProxy frontend and the ORCHESTRATOR_LEADER_URL variable on each node, which an image pull does not add. Nothing breaks without them, and both changes are described in the HA documentation.

Full instructions are available in the ProxCenter update guide.

Thank you to everyone who contributed feedback, testing and issue reports for this release.