Troubles after update to Debian trixie: network interface not found (Exim4 conflict issue)

[daitangio]

After upgrading from Debian bookworm to trixie + a reboot the docker-mailserver was unable to start.

It reported inability to detect proper hostname.
I forced the FQN name (changing the config) and then the error changed: the mailserver cannot find eth0 interface
I got "NETWORK_INTERFACE appears to be misconfigured,"
It seems trixie (or docker?) is hiding eth0 interface inside the container: I got error about system unable to find eth0.

eth0 is visible from host server

I have searched on google and on the documentation on some hints on how to solve this, without any luck.
Had Someone similar problems (with possible solutions)?

I was forced to do a rollback in a hurry to restore system
Below my docker-compose:

services:
  mail:
    # Ref
    # https://github.com/docker-mailserver/docker-mailserver/blob/master/CHANGELOG.md#1120
    image: mailserver/docker-mailserver:15.1.0
    hostname: ${HOSTNAME}
    domainname: ${DOMAINNAME}
    container_name: ${CONTAINER_NAME}
    ports:
    - "25:25"
    - "143:143"
    - "465:465"  # ESMTP (implicit TLS)
    - "587:587"
    - "993:993"
    volumes:
    - /opt/mailserver/maildata:/var/mail
    - /opt/mailserver/mailstate:/var/mail-state
    - /opt/mailserver/maillogs:/var/log/mail
    - /etc/localtime:/etc/localtime:ro
    - ./config/:/tmp/docker-mailserver/${SELINUX_LABEL}
    - /etc/letsencrypt/:/tmp/ssl:ro
    environment:
      - SSL_TYPE=manual
      - SSL_CERT_PATH=/tmp/ssl/live/gioorgi.com/fullchain.pem
      - SSL_KEY_PATH=/tmp/ssl/live/gioorgi.com/privkey.pem
    env_file:
    - .env
    - mailserver.env
    cap_add:
    - NET_ADMIN
    - SYS_PTRACE
    restart: always
    stop_grace_period: 1m
    mem_limit: 800M
    #cpu_percent: 40
    healthcheck:
      # Reference https://docs.docker.com/engine/reference/builder/#healthcheck
      test: "ss --listening --tcp | grep -P 'LISTEN.+:smtp' || exit 1"
      interval: 45s
      timeout: 3s
      retries: 2

[polarathene]

The related logic is here:

docker-mailserver/target/scripts/startup/setup.d/networking.sh

Lines 16 to 27 in 066d314

	CONTAINER_IP=$(ip addr show "${NETWORK_INTERFACE}" | \
	grep 'inet ' | sed 's|[^0-9\.\/]*||g' | cut -d '/' -f 1)
	CONTAINER_NETWORK=$(echo "${CONTAINER_IP}" | cut -d '.' -f1-2).0.0
	if [[ -z ${CONTAINER_IP} ]]; then
	_log 'error' 'Detecting the container IP address failed'
	_dms_panic__misconfigured 'NETWORK_INTERFACE' 'Network Setup [docker_permit]'
	fi
	while read -r IP; do
	CONTAINER_NETWORKS+=("${IP}")
	done < <(ip -o -4 addr show type veth | grep -E -o '[0-9\.]+/[0-9]+')

If this changed from your Debian host upgrade, you'll need to see what the container has for a network interface in Trixie, like this:

# Will list each interface available to the container,
# ignore the loopback interface `lo` (`127.0.0.1`)
docker compose exec mail ip addr show

It's not DMS specific, this would change across any containers (just most containers won't need to know).

I have searched on google and on the documentation on some hints on how to solve this, without any luck.
Had Someone similar problems (with possible solutions)?

Our ENV docs page documents NETWORK_INTERFACE:

So you can set that to match the appropriate interface from the ip addr show command and DMS will use that.

Please do respond back with what the new interface name is for the benefit of others in your situation 👍

---

Feedback about your compose.yaml

It reported inability to detect proper hostname.

You have possible got that misconfigured, but it's hard to tell. This was reworked around DMS v10.3.0 I think (our compose.yaml example wasn't updated to only show hostname usage until April 2023)

    hostname: ${HOSTNAME}
    domainname: ${DOMAINNAME}

This is outdated config (by many years), it should just be hostname, like mail.example.com. We do not have a DOMAINNAME and to my knowledge that's not a standard ENV? Have you set that yourself somewhere, or is Debian exporting it for you somewhere?

Presumably you've got DOMAINNAME inherited from .env which Docker Compose can use by default. You've added that to env_file which forwards that into the DMS container ENV as well, you should not be setting DOMAINNAME (or HOSTNAME) this way in DMS ENV. If the .env is only intended for use in your Docker Compose file, I would remove it from env_file.

    cap_add:
    - NET_ADMIN
    - SYS_PTRACE

You should not need SYS_PTRACE either. Assuming you have Fail2Ban enabled NET_ADMIN is still required though.

You may want to look at your mailserver.env where you may have some ENV to cleanup (For example, ONE_DIR that was removed two years ago). Be mindful of config drag like that when upgrading to new major versions of DMS, especially breaking changes if you're not already in the habit of doing so :)

services:
  mail:
    # Ref
    # https://github.com/docker-mailserver/docker-mailserver/blob/master/CHANGELOG.md#1120
    image: mailserver/docker-mailserver:15.1.0

Unclear what the ref link was regarding in the changelog other than the DMS v11.2 release (Oct 2022) which probably explains why the config is in the state that it is 😓

[polarathene]

So new version of docker can change the network interface?

It could be due to a variety of things. I still see eth0 on some systems in a container with recent versions of Docker Engine, so it's likely specific to Debian Trixie packaging / config somewhere perhaps.

[casperklein]

No problem here: Debian 13 (Trixie) host with Docker version 29.1.5, still gives me eth0.

[daitangio]

Thank you. I am still analyzing the problem.
On Debian Trixie docker still return eth0 as said by @casperklein

I suspect exim4 was installed automatically and it was occupying the same port I wanted to use (!), and this could be the root cause.
I found this reference https://mailcow.email/posts/2025/upgrade-deb12-to-deb13/#removing-exim-after-the-reboot-if-present
I will try to remove exim4 after trixie update.

[daitangio]

I confirm the problem was caused by exim4 keeping port 25 busy. This in turn caused strange error on boot of docker mail.
The issue is that exim do not start until you reboot the machine: so the issue pop ups at first reboot.

Solution:
After upgrading to trixie, ensure exim4 is unistalled issue something with
sudo apt purge exim4 exim4-base exim4-config exim4-daemon-light
or fix the port conflict in some way.