Ir al contenido

Administration and Troubleshooting in Proxmox VE

Advanced administration manual for Proxmox VE, cluster quorum, Ceph, and hybrid deployment of Universal Print in Windows Server laboratory environments.

0views

1. Ceph in Proxmox: Hyperconvergence and High Availability

Sección titulada «1. Ceph en Proxmox: Hiperconvergencia y Alta Disponibilidad»

Ceph is a software-defined storage (SDS) engine integrated into Proxmox VE. It allows for the unification of physical disks (SSD/NVMe) from multiple servers to consolidate a single logical pool of distributed, shared, and redundant storage.

  1. Distributed Storage: Groups dispersed local hard drives, eliminating the need to acquire proprietary external SAN/NAS arrays.
  2. High Availability (HA): If a physical node fails, data remains transparently accessible on the remaining nodes, allowing the HA manager to immediately restart the affected Virtual Machines (VMs).
  3. Live Migration: Allows moving running virtual machines between physical nodes without shutdown or service interruptions, as storage is mounted on a cluster.
  4. Self-Healing: Upon replacing a faulty disk, Ceph autonomously initiates reconstruction and indexing to restore the pool's logical redundancy.

To ensure a stable data quorum in production:

  • Minimum of 3 Nodes: An essential requirement to avoid split-brain conflict scenarios.
  • Dedicated Storage Network: A 10 Gbps physical link exclusive to Ceph replication traffic. 1 Gbps networks will induce high latency and bottlenecks.
  • Solid State Drives (SSD/NVMe): Required to provide the necessary IOPS in dynamic virtualized environments.

2. Changing the Hostname in Proxmox VE

Sección titulada «2. Cambio de Hostname en Proxmox VE»

Proxmox VE uses the terminal hostname (hostname) to structure the physical paths of its read-only cluster file system (/etc/pve/nodes/NOMBRE_NODO). Changing this value directly while running will break connectivity and logical mounting.

From the local SSH console on the server:

  1. Edit the system hostname:
    Ventana de terminal
    nano /etc/hostname
    # Modifica 'nombre-antiguo' por 'PT-PORTEL-ADMIN-NODE01'
  2. Modify the local name resolution in the hosts base:
    Ventana de terminal
    nano /etc/hosts
    # Asegura la IP correspondiente:
    # 192.168.10.50 PT-PORTEL-ADMIN-NODE01.local PT-PORTEL-ADMIN-NODE01

Upon reboot, Proxmox will generate an empty directory structure based on the new hostname. Old VM and LXC configurations must be manually copied to the new logical location:

  1. Restart the physical node:
    Ventana de terminal
    reboot
  2. Log in again via SSH and execute the .conf file migration:
    Ventana de terminal
    cp /etc/pve/nodes/NOMBRE_ANTIGUO/qemu-server/* /etc/pve/nodes/PT-PORTEL-ADMIN-NODE01/qemu-server/
    cp /etc/pve/nodes/NOMBRE_ANTIGUO/lxc/* /etc/pve/nodes/PT-PORTEL-ADMIN-NODE01/lxc/

Once verified via the web GUI that the logical resources and VMs are accessible under the new node, proceed to delete the obsolete remnants on the fly:

Ventana de terminal
rm -rf /etc/pve/nodes/NOMBRE_ANTIGUO

To disconnect a node from a communication cluster (Corosync) and restore its standalone condition without formatting the physical host:

Ventana de terminal
systemctl stop pve-cluster
systemctl stop corosync

The /etc/pve file system is locked in read-only mode if there is no network quorum. To alter it locally:

Ventana de terminal
pmxcfs -l
Ventana de terminal
rm -f /etc/pve/corosync.conf
rm -rf /etc/corosync/*

Step 4: Closing Maintenance Mode and Standard Activation

Sección titulada «Paso 4: Cierre del Modo Mantenimiento y Activación Estándar»
Ventana de terminal
killall pmxcfs
systemctl start pve-cluster

  • Symptom: All virtual machines show icons with gray question marks and it is not possible to edit or start them.
  • Cause: The node has lost communication with the rest of the Corosync cluster and blocks metadata writing to prevent logical collisions.
  • Emergency Bypass (1 Maximum Vote):
    Ventana de terminal
    pvecm expected 1

Error 2: Permission Denied when Deleting Node Folders with Root

Sección titulada «Error 2: Permission Denied al Borrar Carpetas de Nodos con Root»
  • Symptom: You attempt to delete a ghost node with rm -rf /etc/pve/nodes/antiguo and the system rejects the operation at the kernel level.
  • Cause: Loss of quorum blocking the cluster base pmxcfs in read-only mode.
  • Solution: Run the local mount sequence:
    Ventana de terminal
    systemctl stop pve-cluster
    pmxcfs -l
    rm -rf /etc/pve/nodes/antiguo
    killall pmxcfs
    systemctl start pve-cluster

5. Deployment Manual: Hybrid Printing with Universal Print

Sección titulada «5. Manual de Despliegue: Impresión Híbrida con Universal Print»

This case study details the integration architecture of a print gateway on Windows Server Core running under Proxmox VE, to connect local physical printers to the Microsoft 365 Universal Print SaaS service.

[💻 Cliente Windows 11] ───► [☁️ Universal Print] ───► [🔌 Conector (VM Core)]
(PT-PORTEL-ADMIN) (Microsoft 365) (w-server-2022)
│ │
▼ ▼
(Gestión Remota MMC) ────────────────────────────────► [🖨️ Impresoras Red]
  • Host Hypervisor: Proxmox VE node with SSD storage.
  • Gateway Server (VM Guest): Windows Server 2022 Core (Minimal network installation).
    • Server Logical IP: 192.168.10.22
    • Hostname: w-server-2022
  • Administration Client PC: Windows 11 (PT-PORTEL-ADMIN) in the local management subnet, authenticated in Microsoft Entra ID (AzureAD\PortelAdmin).
    • Admin IP: 192.168.10.164

During the Windows Server Core installation startup on Proxmox, no target physical disk will be displayed because the Windows kernel lacks drivers for SCSI controllers optimized for virtualized environments.

  • Resolution:
    1. Ensure the virtio-win.iso image is mounted in the Proxmox virtual CD console.
    2. Click Load Driver in the Windows installer.
    3. Select the physical path: D:\viostor\2k22\amd64.
    4. If the disk appears inactive in the installation terminal, open the console using Shift + F10 and activate it from Diskpart:
      Ventana de terminal
      diskpart
      select disk 0
      attributes disk clear readonly
      online disk
      exit

Since the Windows Core minimal interface does not have a classic graphical console (devmgmt.msc), the virtualized network card drivers must be provisioned via PowerShell commands:

Ventana de terminal
pnputil /add-driver "D:\NetKVM\2k22\amd64\netkvm.inf" /install

Once the NetKVM driver is loaded, run the basic administrative tool sconfig to assign a permanent static IP address on the laboratory server subnet.


5.3. Phase 2: Advanced Remote Administration Configuration (CredSSP and UAC)

Sección titulada «5.3. Fase 2: Configuración Avanzada de Administración Remota (CredSSP y UAC)»

Since our laboratory is located in an isolated Workgroup (without a local Active Directory) and the client PC is integrated into Entra ID cloud (AzureAD\PortelAdmin), it is mandatory to enable CredSSP and delegate explicit remote permissions to allow the Windows 11 RSAT tool to remotely manage the Core server's print subsystem.

Step 1: Change the Core Server Network Profile to Private

Sección titulada «Paso 1: Cambiar el Perfil de Red del Servidor Core a Privado»

If the Windows Server Core network card is configured with the “Public” category, the firewall will block all DCOM/WMI-based remote management connections:

Ventana de terminal
# Identificar el InterfaceIndex del adaptador
Get-NetConnectionProfile
# Reemplazar el número de índice por el correcto obtenido (ejemplo: index 12)
Set-NetConnectionProfile -InterfaceIndex 12 -NetworkCategory Private

In the elevated PowerShell terminal of your Windows 11 management machine (PT-PORTEL-ADMIN), add the target server to the WinRM secure hosts list:

Ventana de terminal
winrm set winrm/config/client '@{TrustedHosts="w-server-2022,192.168.10.22"}'

To authorize the secure transmission of network passwords in a double-hop during print queue provisioning:

  • On the Windows 11 Client PC:
    Ventana de terminal
    Enable-WSManCredSSP -Role Client -DelegateComputer "w-server-2022"
  • On the Core Print Server:
    Ventana de terminal
    Enable-WSManCredSSP -Role Server

By default, Windows applies a strict policy that downgrades the administrative privileges of local accounts over remote WMI connections. Run this command in the Windows Server Core terminal to disable it:

Ventana de terminal
New-ItemProperty -Name LocalAccountTokenFilterPolicy -Path "HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System" -Value 1 -PropertyType DWord -Force

The Universal Print Connector requires the OS basic queue framework and infrastructure to be installed on the target node:

Ventana de terminal
Install-WindowsFeature Print-Services -IncludeManagementTools

5.4. Phase 3: Remote Connection and Express Printer Duplication (Point and Print)

Sección titulada «5.4. Fase 3: Conexión Remota y Duplicación Expresar de Impresoras (Point and Print)»

To allow Ricoh or Zebra drivers to be imported and installed remotely on the VM without requiring physical UAC authorizations:

Ventana de terminal
$regPath = "HKLM:\SOFTWARE\Policies\Microsoft\Windows NT\Printers\PointAndPrint"
if (-not (Test-Path $regPath)) { New-Item -Path $regPath -Force }
Set-ItemProperty -Path $regPath -Name "Restricted" -Value 0 -Type DWord -Force
Set-ItemProperty -Path $regPath -Name "NoWarningNoElevationOnInstall" -Value 1 -Type DWord -Force

Due to a recurring Windows issue when managing remote network authentications from Azure AD-joined clients to standalone local servers, you must initialize the remote print administration console (printmanagement.msc) by forcing the server's credential scope using the command runas /netonly:

Ventana de terminal
# Ejecuta en el CMD de tu PC de administración (PT-PORTEL-ADMIN)
runas /netonly /user:w-server-2022\Administrador "mmc printmanagement.msc"

This command will launch the Windows 11 remote management console using your local identity, but the communication within the subnet to the VM w-server-2022 will be implicitly authorized with the server's administrative role, allowing printers and TCP ports to be added seamlessly.

Instead of adding the 10 network printers one by one on the Core server:

  1. Export: In the MMC graphical console opened with the previous command, right-click on the local administrator PC node (PT-PORTEL-ADMIN) and select Export printers to a file….
  2. Import: In the same console tree, right-click on the remote server (w-server-2022) and select Import printers from a file…, loading the .printerExport file to migrate the 10 ports, configurations, and drivers in less than a minute.

5.5. Phase 4: Final Registration in Universal Print

Sección titulada «5.5. Fase 4: Registro Final en Universal Print»

Once the local queues are configured on the virtual server, open a remote PowerShell session on the Core server and run the connector to enroll your print resources in the cloud under your authorized Microsoft 365 organization:

Ventana de terminal
# Instalación del paquete del conector en modo desatendido
C:\Deploy\UniversalPrintConnector.exe /quiet ACCEPT_EULA=Y AUTH_CODE="CODIGO_REGISTRO_ENTRA_ID"

  • ✓ Quórum Garantizado: Servidores de clúster Proxmox resolviendo comunicaciones y enrutando almacenamiento de forma redundante.
  • ✓ Seguridad Híbrida: Enlaces remotos WinRM asegurados bajo CredSSP y UAC filtrado para administración de impresión.
  • ✓ Cuentas Limpias: Identidades y credenciales de red simuladas en consonancia con la arquitectura lógica de laboratorio de Portelcore.