Proxmox Upgrades

Routine updates are one thing. A Proxmox point release with a new kernel is another creature entirely.

This section is for the higher-risk path: the work that starts with a backup window, touches the host kernel, crosses GPU driver boundaries, and can expose old assumptions in networking, initramfs, or hardware drivers.

The current story has a very specific order:

PVE 9.2 upgrade
  -> kernel 7.0 becomes the default
  -> reboot exposes a nova_core + ifupdown2-pre boot hang
  -> emergency shell diagnosis
  -> nova_core blacklist and vmbr0 repair
  -> host returns on 192.168.50.20

The boot hang is not a separate incident. It is the side effect that came into view after the PVE 9.2 upgrade.

In This Section

• PVE 9.2 Upgrade Runbook - the pre-flight checks, PBS backup window, container package updates, host upgrade, NVIDIA 595.71.05 driver sync, verification, cleanup, and recovery paths.
• Kernel 7 Boot Hang RCA - the post-upgrade failure where kernel 7.0.6-2-pve auto-loaded nova_core, blocked udevadm settle, and exposed a broken eth0/missing vmbr0 network config.

How To Read It

Read the PVE 9.2 runbook first. It is the planned work.

Read the RCA second. It is what the reboot taught the machine.

If you are not doing a major upgrade yet, start with Update And Maintenance. That page covers the normal maintenance rhythm. Come back here when the operation includes a new kernel, DKMS, NVIDIA driver changes, or network-interface risk.

Related Topics

• Proxmox Backup Server - the backup safety net used before the upgrade.
• GPU Passthrough On Proxmox - the broader host/container driver model behind the NVIDIA steps.
• Storage And ZFS - the ZFS and snapshot layer underneath rollback planning.