When creating an LXC container in Proxmox 9, /run and /tmp are not mounted as tmpfs.

This causes e.g. the eth0 network interface to stay DOWN after

pct start

(on proxmox), even though

/etc/network/interfaces

is correctly configured with

allow-hotplug eth0

(

auto eth0

has same behavior) and the right IP.

**Workaround:** After boot, running

ifdown eth0 && ifup eth0

brings the interface up correctly or removing some files in /run at early stages:

/run/network/ifstate /run/network/ifstate.* /run/network/.ifstate.lock

Reproduction

Reproducible for both

1.  devuan-6.0-excalibur_20251124_amd64.tar.xz https://fra1lxdmirror01.do.letsbuildthe … tfs.tar.xz 29adc79fae8c0a0be17533fbe91b85d87fe0133753572575b111479c056dccd4
2.. Devuan Excalibur LXC image from linuxcontainers.org (2026-02-20)

pct create 130 local:vztmpl/devuan-6.0-ansible-ready.tar.zst \
  --hostname test --storage local-zfs --cores 2 --memory 8192 --swap 8192 \
  --rootfs local-zfs:80 --net0 "name=eth0,bridge=vmbr1,ip=10.0.0.30/24,gw=10.0.0.254" \
  --unprivileged 1 --onboot 1

pct start 130
pct exec 130 -- ip addr show eth0
# Shows: state DOWN, qdisc noop, no inet address

Current Status

What doesn't work is mounting /run within the bootup, as we want the container to not have these privileges.
The mount already present in the bootup scripts fails silently.

What we are currently investigating if

pct --features ,mount=<fstype;fstype;....

can be helpful.
https://pve.proxmox.com/pve-docs/pct.1.html

Does anyone use Devuan Excalibur as container in Proxmox and can us help here?