ProxCenter v1.4.12 Release

ProxCenter v1.4.12 is a feature release focused on migration, data protection, operations and automation. It introduces backup restore drills, alert notifications to Slack, Teams and webhooks, host drain, per-VM failover in Site Recovery, write access for API tokens, and warm migrations that convert the guest at cutover.

---

🚚 Migration

Conversion at cutover

Warm migrations now convert the guest at cutover: the VM boots on VirtIO and keeps its static IP, for both VMware and XCP-ng sources.

A new migration dialog

The migration dialog is redesigned in two columns with a review step. It sets the network per NIC, overrides hardware settings, picks a target storage per disk, and lets some disks be left behind.

Safer and faster at scale

• Bulk migrations run in parallel from a server-side queue, and warm copies can use a chosen VMkernel network
• Every copied disk is verified before a VMware VM is handed over, and a vTPM source keeps its TPM
• Thick zeroing is capped per target cluster, and leftover nbdkit readers are swept
• A failed migration shows the reason Proxmox gives, and a guest is reported migrated only once its task succeeded
• A storage plugin warning no longer breaks warm migrations, and a failed warm job cleans up after itself
• "Cutover now" stays reachable when the migrate dialog is minimized

---

💾 Backups and Site Recovery

Backup restore drills

ProxCenter can now test-restore PBS backups in an isolated VM, on demand or on a schedule, to prove that backups can actually be restored.

Per-VM failover

A single VM of a DR plan can now be failed over and failed back on its own.

Backup coverage and restores

• Guests covered by no backup job are listed and raise an alert
• Files can be restored straight into a running guest, limited to the tenant's own vDC backups
• PBS backups are browsed by machine, and restore points open in a detail dialog
• A full restore keeps the MAC addresses and overwrites the existing VM, and can leave the backed-up tags behind

Replication

Replication jobs are now editable, and an interrupted or cancelled sync puts the job back on schedule. Replication snapshots are named in UTC and ordered by creation, so the end of daylight saving time no longer breaks the next sync.

---

🖥️ DRS, nodes and HA

Drain a host

A host can now be drained from the inventory: maintenance mode, migration of all VMs and reboot are handled through one evacuation plan that refuses what cannot move.

Clearer placement

DRS explains placement per guest and why an action is refused. A rolling update waits until the node is really evacuated, DRS stays out of the cluster meanwhile, and a node in maintenance no longer reads as offline.

Nodes and SSH

• Each node's SSH address and port are configurable, and used on every SSH path
• An SSH command no longer runs twice when the orchestrator answer is lost
• A Proxmox connection is checked with read-only probes, with a fix suggested for each failure
• The HA control plane converts on air-gapped sites and from a private registry, and the HTTP proxy reaches the orchestrator container
• A cluster can be excluded from the license quota, with the overflow shown per license

---

🔔 Alert notifications

Alerts can now be sent to Slack, Microsoft Teams, ntfy, Discord and webhooks.

The bell, the summary and the alerts page count and list the same alerts, and a new Top consumers widget ranks guests by CPU, memory, disk or network over a time window, with a metrics retention set from the interface.

Network Flows tells sFlow reception apart from configuration, and alerts when a configured node stops reporting.

---

🔐 Security and access

API tokens can write

API tokens can now write, with per-creator rights, tenant confinement and audit.

Sign-in

• The login lockout policy is enforced, and client IPs are resolved through trusted proxies
• LDAP sign-in refusals are explained through a connection test and a read-only user lookup
• Every role a user cumulates from SSO is shown, and tenant view saves no longer erase them
• The Events view shows which ProxCenter user opened a console

---

🏢 Multi-tenancy and vDC

• vDC usage history: sampler, retention, charts and CSV export, deleted vDCs included
• Tenants can create pure L2 networks, the VNI choice is scoped to the transport, and the L3 edge model is explicit per vDC
• Catalogue images and blueprints are granted per tenant or per vDC, and the grant is enforced at deploy time
• A provider can deploy a template into a tenant vDC, and tenant template images are staged on the vDC ISO library
• Storage rights are scoped to named storages for ISO uploads and backup destinations

---

✨ Also in this release

Area	What's new
🔥 Firewall	Logs searched and filtered per security group, rule, node and cluster; guest firewall state and default policies shown as Proxmox applies them
🗂️ Pools	Proxmox pools created, deleted and commented from the Pools view
🏷️ Inventory	Guests found by tag in the inventory tree, sortable Tags column, each Ceph OSD shown with its own state
📄 Reports	VMs and containers told apart; the inventory report lists storages and guest networks
📈 Metrics	A guest's total disk size shown and metered; OpenTelemetry metric servers save with the options PVE expects
📤 Uploads	Storage upload chunks retried after a proxy timeout, without being written twice
📦 Dependencies	Next.js, sharp, msw 3, vitest 5 and the patch groups updated; zlib patched in both images

---

⬆️ Upgrading

Fifteen database migrations are included, all additive. They run automatically at first boot, and the orchestrator updates its own schema when it starts.

docker compose pull && docker compose up -d

Three points deserve attention:

• The metrics history is now pruned: the orchestrator keeps 14 days by default and removes older samples. To keep more, raise the retention in Settings > Alert thresholds right after the upgrade.
• Custom roles and user grants holding vm.view also receive storage.content, which the guest views need.
• HA clusters converted before 1.4.12 keep their compose file: to pull from a private registry or go through an HTTP proxy, add REGISTRY, HTTP_PROXY, HTTPS_PROXY and NO_PROXY to /opt/proxcenter/.env and to the compose file on each node. Nothing changes for a cluster that needs neither.

Air-gapped sites upgrade from the bundle: the proxcenter-community-1.4.12.tar.gz asset for Community, or the Air-gapped tab of your proxcenter.io account for Enterprise, as described in the air-gapped installation guide.

Full instructions are available in the ProxCenter update guide.

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