xenial: lxcbr0 does not initiate correctly after reboot

[Michael Gugino]

After reboot, lxcbr0 is not initiated correctly. The interface is in some unusable but configured state. The interface is usable after running ifdown lxcbr0; ifup lxcbr0. However, ifdown fails with some output, as seen below:

<system reboot>
root@1604aio:~# brctl show
bridge name bridge id STP enabled interfaces
br-mgmt 8000.000000000000 no
br-storage 8000.000000000000 no
br-vlan 8000.d6f72b3e87b4 no br-vlan-veth
br-vxlan 8000.000000000000 no

root@1604aio:~# ifup lxcbr0
ifup: interface lxcbr0 already configured

root@1604aio:~# brctl show
bridge name bridge id STP enabled interfaces
br-mgmt 8000.000000000000 no
br-storage 8000.000000000000 no
br-vlan 8000.d6f72b3e87b4 no br-vlan-veth
br-vxlan 8000.000000000000 no

root@1604aio:~# ifdown lxcbr0
Cannot find device "lxcbr0"
Cannot find device "lxcbr0"
Cannot find device "lxcbr0"
run-parts: /etc/network/if-post-down.d/bridge exited with return code 1
Removing LXC IPtables rules.
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: Bad rule (does a matching rule exist in that chain?).
iptables: No chain/target/match by that name.
iptables: No chain/target/match by that name.
LXC IPtables rules removed.

root@1604aio:~# ifup lxcbr0
Creating LXC IPtables rules.
LXC IPtables rules created.
Starting LXC dnsmasq.
dnsmasq started.
root@1604aio:~# brctl show
bridge name bridge id STP enabled interfaces
br-mgmt 8000.000000000000 no
br-storage 8000.000000000000 no
br-vlan 8000.d6f72b3e87b4 no br-vlan-veth
br-vxlan 8000.000000000000 no
lxcbr0 8000.000000000000 no

[Jimmy McCrory (jimmy-mccrory)]

Looks like this is caused by a conflict with the lxc-net service.

May 11 21:01:00 host2-osa dnsmasq[1100]: failed to create listening socket for 10.100.100.1: Cannot assign requested address
May 11 21:01:00 host2-osa dnsmasq[1100]: FAILED to start up
May 11 21:01:00 host2-osa lxc-net[994]: Failed to setup lxc-net.

In 14.04 the service was disabled, it doesn't seem to be disabled in systemd systems.

The options would be to either disable lxc-net, or to remove /etc/network/interfaces.d/lxc-net-bridge.cfg and let lxc-net handle managing the bridge.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/356677
Committed: https://git.openstack.org/cgit/openstack/openstack-ansible-lxc_hosts/commit/?id=196807bbda5f7c2dbafff9270b43a571956487f0
Submitter: Jenkins
Branch: master

commit 196807bbda5f7c2dbafff9270b43a571956487f0
Author: Michael Gugino <email address hidden>
Date: Wed Aug 17 15:49:25 2016 -0400

    Mask lxc-net.service for systemd

    On systemd-based systems, lxc-net is started whenever
    lxc is started. This is because lxc-net is declared
    as a dependency of lxc.

    We currently mask the service on upstart systems.
    We are presently manually configuring the lxcbr0
    on all systems, systemd-based systems included.

    This patch will mask the lxc-net.service to enable
    lxcbr0 to be brought up without error due to conflicting
    lxc-net.service.

    Closes-Bug: #1580766
    Change-Id: I099cf638bbb8874ca45b815383de8588fecbe7b6

[Doug Hellmann (doug-hellmann)]

This issue was fixed in the openstack/openstack-ansible-lxc_hosts 14.0.0.0b3 development milestone.