Centos 7 disable predictable network interface names with packer/vagrant

Question

I'm trying to upgrade our local dev vagrant box to CentOS 7.2 (from 6.8), but ran into a problem with the new "predictable network interface names". My puppet config is expecting eth0 and eth1, but it's getting enp0s3 and enp0s8.

I managed to disable predictable network interface names in the kickstart file by adding:

bootloader --location=mbr --append="net.ifnames=0"

and removing the package biosdevname

Now when my vagrant box boots it has eth0 and eth1 (shown when I do an ip -a), but I don't have the network scripts in /etc/sysconfig/network-scripts/ (only ifcfg-enp0s3 and ifcfg-lo).

When vagrant boots this VM it shows this error:

The following SSH command responded with a non-zero exit status.
Vagrant assumes that this means the command failed!

# Update sysconfig
sed -i 's/\(HOSTNAME=\).*/\1vm.example.com/' /etc/sysconfig/network

# Update DNS
sed -i 's/\(DHCP_HOSTNAME=\).*/\1"vm"/' /etc/sysconfig/network-scripts/ifcfg-*

# Set the hostname - use hostnamectl if available
echo 'vm.example.com' > /etc/hostname
if command -v hostnamectl; then
  hostnamectl set-hostname --static 'vm.example.com'
  hostnamectl set-hostname --transient 'vm.example.com'
else
  hostname -F /etc/hostname
fi

# Remove comments and blank lines from /etc/hosts
sed -i'' -e 's/#.*$//' -e '/^$/d' /etc/hosts

# Prepend ourselves to /etc/hosts
grep -w 'vm.example.com' /etc/hosts || {
  sed -i'' '1i 127.0.0.1\tvm.example.com\tvm' /etc/hosts
}

# Restart network
service network restart


Stdout from the command:

/bin/hostnamectl
Restarting network (via systemctl):  [FAILED]


Stderr from the command:

Job for network.service failed because the control process exited with error code. See "systemctl status network.service" and "journalctl -xe" for details.

journalctl -xe shows:

-- Unit network.service has begun starting up.
Oct 11 04:28:59 vm.example.com network[3130]: Bringing up loopback interface:  Could not load file '/etc/sysconfig/network-scripts/ifcfg-lo'
Oct 11 04:28:59 vm.example.com network[3130]: Could not load file '/etc/sysconfig/network-scripts/ifcfg-lo'
Oct 11 04:28:59 vm.example.com network[3130]: Could not load file '/etc/sysconfig/network-scripts/ifcfg-lo'
Oct 11 04:28:59 vm.example.com network[3130]: Could not load file '/etc/sysconfig/network-scripts/ifcfg-lo'
Oct 11 04:28:59 vm.example.com network[3130]: [  OK  ]
Oct 11 04:28:59 vm.example.com network[3130]: Bringing up interface enp0s3:  Error: Connection activation failed: No suitable device found for this connection.
Oct 11 04:28:59 vm.example.com network[3130]: [FAILED]
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com network[3130]: RTNETLINK answers: File exists
Oct 11 04:28:59 vm.example.com systemd[1]: network.service: control process exited, code=exited status=1
Oct 11 04:28:59 vm.example.com systemd[1]: Failed to start LSB: Bring up/down networking.

How can I keep eth0 and eth1, but make this work correctly?

Thanks

Answer 1

I added a provisioner script in packer that seems to have fixed this problem:

#!/bin/bash

mv /etc/sysconfig/network-scripts/ifcfg-enp0s3 /etc/sysconfig/network-scripts/ifcfg-eth0
sed -i -e 's/enp0s3/eth0/' /etc/sysconfig/network-scripts/ifcfg-eth0
bash -c 'echo NM_CONTROLLED=\"no\" >> /etc/sysconfig/network-scripts/ifcfg-eth0'

Answer 2

How can I keep eth0 and eth1, but make this work correctly?

I believe I can help answer the first part about keeping eth0 and eth1 (or at least get you a good reference). According to Consistent Network Device Naming in Linux, you are supposed to be able to disable the new naming with the following from Section 9 of the manual:

Disable during install time

To disable the use of the new naming scheme, during installation (attended or automated), pass the kernel command line parameter biosdevname=0 on the boot command line. The parameter should be passed on the boot command line after installation to ensure that a new network adapter plugged in post installation has a traditional "eth" name.

Its been my experience its hit-or-miss whether it actually works. See, for example, Network device naming problems and aliases for p2p1 and p3p1 back to eth0 and eth1 on Super User.