sudo restart lxc-net does not kill/restart dnsmasq

[Scott Moser]

$ ps axw | grep dnsm
14672 ? S 0:00 dnsmasq -u lxc-dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/lxc/dnsmasq.pid --conf-file= --listen-address 10.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0

$ sudo vi /etc/init/lxc-net.conf
# change parameters for dnsmasq

$ sudo restart lxc-net
$ ps axw | grep dnsm
14672 ? S 0:00 dnsmasq -u lxc-dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/lxc/dnsmasq.pid --conf-file= --listen-address 10.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0

^^ Note, restart did not get new dnsmasq pid or settings

$ sudo stop lxc-net
lxc-net stop/waiting
$ sudo start lxc-net
lxc-net start/running
$ ps axw | grep dnsm
14672 ? S 0:00 dnsmasq -u lxc-dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/lxc/dnsmasq.pid --conf-file= --listen-address 10.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0

## Note stop, start did not get new dnsmasq settings

$ sudo stop lxc-net
$ sudo kill 14672
$ sudo start lxc-net
lxc-net start/running
$ ps axw | grep dnsm
32299 pts/0 S+ 0:00 grep --color=auto dnsm

## Note, no dnsmasq was started

$ sudo reboot
## on reboot new settings will take affect

ProblemType: Bug
DistroRelease: Ubuntu 12.04
Package: lxc 0.7.5-3ubuntu62
ProcVersionSignature: User Name 3.2.0-29.46-virtual 3.2.24
Uname: Linux 3.2.0-29-virtual x86_64
ApportVersion: 2.0.1-0ubuntu12
Architecture: amd64
Date: Thu Aug 30 01:04:46 2012
Ec2AMI: ami-0000013e
Ec2AMIManifest: FIXME
Ec2AvailabilityZone: nova
Ec2InstanceType: m1.small
Ec2Kernel: unavailable
Ec2Ramdisk: unavailable
ProcEnviron:
 TERM=screen
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: lxc
UpgradeStatus: No upgrade log present (probably fresh install)
modified.conffile..etc.init.lxc.net.conf: [modified]
mtime.conffile..etc.init.lxc.net.conf: 2012-08-30T01:03:44.536628

[James Page (james-page)]

Scott

I can reproduce the restart issue, but using start/stop dnsmasq was restarted:

ubuntu@server-15281:~$ ps axw | grep dnsm
 5367 ? S 0:00 dnsmasq -u lxc-dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/lxc/dnsmasq.pid --conf-file= --listen-address 10.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0
ubuntu@server-15281:~$ sudo stop lxc-net
lxc-net stop/waiting
ubuntu@server-15281:~$ ps axw | grep dnsm
 5409 pts/0 S+ 0:00 grep --color=auto dnsm
ubuntu@server-15281:~$ sudo start lxc-net
lxc-net start/running
ubuntu@server-15281:~$ ps axw | grep dnsm
 5425 ? S 0:00 dnsmasq -u lxc-dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/lxc/dnsmasq.pid --conf-file= --listen-address 10.0.3.1 --dhcp-range 10.0.3.2,10.0.3.254 --dhcp-lease-max=253 --dhcp-no-override --except-interface=lo --interface=lxcbr0

I did get confused by the duplicated configuration entries for network ranges in /etc/default/lxc and /etc/init/lxc-net.conf.

Marking 'Low' as I think it only impacts the restart case - which can be worked around using stop/start lxc-net.

[Serge Hallyn (serge-hallyn)]

Thanks for reporting this bug. This is actually a bug in upstart, not lxc, reported as bug 703800. I'll mark it as a duplicate.

[John Neffenger (jgneff)]

In my tests on Ubuntu 14.04, the LXC dnsmasq process is restarted with:

$ sudo restart lxc-net

only when all guest containers are stopped. If one or more containers are running, the dnsmasq process is left running with its old settings when lxc-net is restarted.

[Serge Hallyn (serge-hallyn)]

This happens because 'restart service' does not cause service's post-stop
to be executed. Only it's pre-stop is executed.

Note that if you want to stop the dnsmasq, you can work around this by
doing

stop lxc-net; start lxc-net

The pre-start also would not be executed on a restart, so just changing
the post-stop would not suffice. We'd have to change the pre-start to
'script'.

I'm not sure whether there is any downside to doing so.

 status: confirmed
 importance: medium

[Pablo RUTH (d-contact-a)]

@jneff yes the same for me, so the workaround doesn't work... because the post-stop script try do delete the bridge but there is an error "bridge lxcbr0 is still up; can't delete it" if all containers are not stopped. After that if you try so start lxc-net there is a test "if [ -d /sys/class/net/${LXC_BRIDGE} ]; then" so as the bridge is already up the script exit and will not launch dnsmasq... I'm right?