[Ubuntu 26.04] Unable to rename interfaces: [busy] Error renaming from ens18 to eth0 with netplan 1.2

[azabrha]

Bug report

cloud-init fails to rename network interface from predictable name (ens18) to desired name (eth0).
The rename operation fails with [busy] error, causing the full network configuration (including DNS, routes, etc.) to not be applied correctly. User usually needs to run netplan apply manually after first boot.

It seems this issue is related to stricter "online" definition in newer netplan (1.2+) as described here:
https://discourse.ubuntu.com/t/spec-definition-of-an-online-system/27838

Steps to reproduce the problem

1. Deploy a fresh Ubuntu 26.04 cloud image on Proxmox (or KVM).
2. Use the following netplan configuration (typical for Proxmox):

network:
  version: 2
  ethernets:
    eth0:
      match:
        macaddress: "xx:xx:xx:xx:xx:xx"
      set-name: eth0
      addresses: [...]
      nameservers: [...]
      routes: [...]

3. Boot the VM.

Environment details

• Cloud-init version: 26.1-0ubuntu2
• Operating System Distribution: Ubuntu 26.04 (Resolute)
• netplan version: 1.2+
• Cloud provider, platform or installer type: Proxmox VE (KVM)

What we tried (did not solve the issue)

• Added ExecStartPre=/bin/sleep (4s, 7s, 10s) on:
  • cloud-init-local.service (/etc/systemd/system/cloud-init-local.service.d/99-delay.conf)
  • cloud-init-network.service (/etc/systemd/system/cloud-init-network.service.d/99-delay.conf)
• Overrode systemd-networkd-wait-online.service (/etc/systemd/system/systemd-networkd-wait-online.service.d/99-override.conf) with:
  • --any --timeout=30/40/60/90 -o routable -i eth0
• Various combinations of delay + timeout override

None of the above reliably fixed the rename failure.

cloud-init logs

2026-05-16 08:12:04,428 - net[WARNING]: Unable to rename interfaces: [['bc:24:11:56:41:9f', 'eth0', None, None]] due to errors: ['[busy] Error renaming mac=bc:24:11:56:41:9f from ens18 to eth0']
2026-05-16 08:12:04,428 - stages.py[WARNING]: Failed to apply network config names: [busy] Error renaming mac=bc:24:11:56:41:9f from ens18 to eth0

(Full log available if needed)

[contributorr]

Hi, yes, this is a really annoying bug that in my case causes VMs to get DHCP and if they do not get rebooted upon first cloud init (because of no new kernel), a static network configuration is not applied until reset. In specific case, when template is older (with older kernel), upon reboot a VM gets proper static configuration, however it takes 2 minutes of waiting for network on first boot nevertheless.