LSB job for "raise network interfaces" times out but configuration works

[Ryan Harper]

On Mon, Mar 26, 2018 at 10:38 AM, Leroy Tennison <leroy at datavoiceint.com> wrote:
> I’ve found numerous posts on the web about this but all solutions fix the
> symptom, not the issue (adjust the networking timeout).  Admittedly, the
> situation is more complex (bonding and bridging) but I would prefer
> something other than a work-around on production hypervisors.  The
> configuration:
>
>
>
> /etdc/modprobe.d/bonding.conf:
>
>                options bonding mode=1 miimon=100 downdelay=200 updelay=200
>
>
>
> /etc/network/interfaces:
>
> source /etc/network/interfaces.d/*          (empty directory)
>
>
>
> # The loopback network interface
>
> auto lo
>
> iface lo inet loopback
>
>
>
> # The primary network interface
>
> auto bond0 br0 enp2s0 enp3s0
>
>
>
> iface enp2s0 inet manual
>
>         bond-master     bond0
>
>
>
> iface enp3s0 inet manual
>
>         bond-master     bond0
>
>
>
> iface bond0 inet manual
>
>         bond-primary    enp2s0 enp3s0
>
>         post-up         ifenslave bond0 enp2s0 enp3s0
>
>         pre-down        ifenslave -d bond0 enp2s0 enp3s0
>
>
>
> iface br0 inet static
>
>         address 10.222.110.95/24
>
>         gateway 10.222.110.1
>
>         up ip route add default via 10.222.110.1 dev br0
>
> #       dns-nameservers 10.222.110.1
>
>         dns-search ofc.datavoiceint.com colo.datavoiceint.com
> client.datavoiceint.com wifi.datavoiceint.com
>
>         bridge_ports    bond0
>
>         bridge stp      off
>
>         bridge_fd       0
>
>         bridge_maxwait  0
>
>
>
> Because of the timeout, the DNS entries in /etc/network/interfaces aren’t
> applied to the system.  I have worked around this by using
> /etc/resolvconf/resolv.conf.d/tail.
>
>
>
> Questions:
>
>                Why is this happening?
>
>                Is there anything I can do to prevent it while maintaining
> the bonding/bridging configuration?

This may help:

Ensure you've bridge-utils and ifenslave packages installed, and then
I suggest this
restructuring of the interfaces file:

auto lo
iface lo inet loopback
    dns-nameservers 10.222.110.1
    dns-search ofc.datavoiceint.com colo.datavoiceint.com
client.datavoiceint.com wifi.datavoiceint.com

auto enp2s0
iface enp2s0 inet manual
    bond-master bond0
    bond-mode active-backup
    bond-primary enp2s0

auto enp3s0
iface enp3s0 inet manual
    bond-master bond0
    bond-mode active-backup
    bond-primary enp2s0

auto bond0
iface bond0 inet manual
    bond-mode active-backup
    bond-slaves none
    bond-primary enp2s0

auto br0
iface br0 inet static
    address 10.222.110.95/24
    gateway 10.222.110.1
    bridge_ports bond0
    bridge_stp off


The changes I've made here are

1) drop up-down hooks, the bridge-utils/ifenslave packages already
provide scripts to do this
2) set mode to active-backup, I suspect with bond-primary this is what you want
3) bond-primary takes a single interface, this interface is always
preferred over any other, in active-backup, it'll fail over to any
other interface that's been enslaved
    I'm not aware of bond-primary accepting a list in case you've
enslaved more than two interfaces
4) make use of bond-slaves none to deal with parallel ifup on physical
interfaces
5) best-practices for bonding replicate the bond configration on
slaves in case one fails to come up, the config is always present
6) drop use of bridge_maxwait as you don't have an anonymous bridge
(no ip address) so you must wait for the underlying device (bond0) to
come up

I'm not entirely sure why resolveconf wasn't picking up the dns-*
settings from the bridge, but I do think that the maxwait 0 mean that
the bridge
wasn't waiting for the underlying interface (bond0) to be come
available.  In general it's going to take non-zero amount of time for
the bond to become
active after the interfaces are cold-plugged; so I think it's worth
letting the ifupdown hook scripts run with the defaults to see if that
gets everything
configured.  If it comes up fine, then from there you can see about
tweaking settings to lower the amount of time it's waiting.

Hope that helps,
Ryan