RTNETLINK answers: File exists after adding ipv6 address

Question

I've been pounding my head against the wall for a while on this and haven't made any headway.

I have a system with a statically assigned ipv4 and ipv6 address. After boot running systemctl status networking results in this:

● networking.service - Raise network interfaces
   Loaded: loaded (/lib/systemd/system/networking.service; enabled; vendor preset: enabled)
  Drop-In: /run/systemd/generator/networking.service.d
       └─50-insserv.conf-$network.conf
   Active: failed (Result: exit-code) since Sat 2016-08-27 14:48:50 MST; 8min ago
     Docs: man:interfaces(5)
  Process: 3301 ExecStart=/sbin/ifup -a --read-environment (code=exited, status=1/FAILURE)
  Process: 3275 ExecStartPre=/bin/sh -c [ "$CONFIGURE_INTERFACES" != "no" ] && [ -n "$(ifquery --read-environment --list --exclude=lo)" ] && udevadm settle (code=exited, status=0/SUCCESS)
 Main PID: 3301 (code=exited, status=1/FAILURE)

Aug 27 14:48:25 phoenix systemd[1]: Starting Raise network interfaces...
Aug 27 14:48:28 phoenix ifup[3301]: /sbin/ifup: waiting for lock on /run/network/ifstate.ens160
Aug 27 14:48:50 phoenix ifup[3301]: RTNETLINK answers: File exists
Aug 27 14:48:50 phoenix ifup[3301]: Failed to bring up ens160.
Aug 27 14:48:50 phoenix systemd[1]: networking.service: Main process exited, code=exited, status=1/FAILURE
Aug 27 14:48:50 phoenix systemd[1]: Failed to start Raise network interfaces.
Aug 27 14:48:50 phoenix systemd[1]: networking.service: Unit entered failed state.
Aug 27 14:48:50 phoenix systemd[1]: networking.service: Failed with result 'exit-code'.

But the interface is actually up and functioning properly.

If I run systemctl restart networking, it fails. Running ifdown ens160 just says the interface is not configured.

If I force the interface down with ifdown --force ens160 it goes down and will come back up with either ifup ens160 or systemctl restart networking.

If I comment out the ipv6 section it's fine after booting.

Here's my interfaces file:

# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

source /etc/network/interfaces.d/*

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto ens160
iface ens160 inet static
    address XXX.XXX.XXX.3/24
    gateway XXX.XXX.XXX.1
    dns-search example.org
    dns-nameservers ::1 127.0.0.1

# This is an IPv6 interface
iface ens160 inet6 static
    address XXXX:XXXX:XXXX:XXXX::3/64
    gateway XXXX:XXXX:XXXX:XXXX::1

This is with a vanilla Ubuntu 16.04.1 server installation on a vmware cluster with only bind9 and ssh server installed. The only configuration changes made are a properly configured bind9 and the static IP changes.

I have more than one installation doing the same thing.

I've done plenty of searching and have come up with nothing matching this situation. My google-fu is usually strong, but it has failed me this time.

Any help I could get would be much appreciated.

Answer 1

Since it works when you comment out the gateway line, you suffer from an unfortunate race condition: As soon as the link of the interface is up, Linux starts doing Neighbor Discovery and accepting Router Advertisements, which can put IPv6 routes into your routing table although the interface is not fully configured yet. When the script later tries to add the default route you get

RTNETLINK answers: File exists

To work around this, you can

• either comment out the gateway line (which is the recommended way provided that routes get announced correctly in your network)
• or disable the acceptance of RAs via

  sysctl -w net.ipv6.conf.device.accept_ra=0
  

  (device being an actual device, default or all).