Excessive dnsmasq-dhcp log entries

[Matt Thompson]

I'm seeing a heap of the following in /var/log/syslog on my controller nodes:

May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_keystone_container-23342668 to the DHCP lease of 10.0.3.153 because the name exists in /etc/hosts with address 172.29.237.189
May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_horizon_container-03ea2807 to the DHCP lease of 10.0.3.192 because the name exists in /etc/hosts with address 172.29.239.184
May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_cinder_api_container-f2301af2 to the DHCP lease of 10.0.3.111 because the name exists in /etc/hosts with address 172.29.239.216
May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_galera_container-874a3e04 to the DHCP lease of 10.0.3.85 because the name exists in /etc/hosts with address 172.29.237.245
May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_nova_conductor_container-832856c3 to the DHCP lease of 10.0.3.145 because the name exists in /etc/hosts with address 172.29.238.175
May 19 15:56:22 jenk-heat-247-node1 dnsmasq-dhcp[28643]: not giving name jenk-heat-247-node1_nova_scheduler_container-70ba2ef0 to the DHCP lease of 10.0.3.200 because the name exists in /etc/hosts with address 172.29.238.236

We could pass LXC's dnsmasq --no-hosts, which will prevent dnsmasq from reading /etc/hosts.

[Serge van Ginderachter (svg)]

On a side note, why do we need this network? the br-mgmt network should suffice, imho.

[Kevin Carter (kevin-carter)]

The fix needs to be added into the the lxc-system-manage tool which is used to bring the dnsmasq oinline for use within lxc

Master:
https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/lxc_hosts/templates/lxc-system-manage.j2#L141-L158

The same thing can be done in Juno too:
https://github.com/stackforge/os-ansible-deployment/blob/juno/rpc_deployment/roles/lxc_common/files/lxc-system-manage#L134-L150

@Serge
As for the lxcbr0 interface, this is the device that all of the traffic leaving the lxc containers goes through. In master / Kilo this devices could be overriden and or changed using the `lxc_net_bridge` variable. There are several options available for setting up and defining the lxc network interface and will need to be configured for a successful deployment when using physical devices. These options can all be seen here: https://github.com/stackforge/os-ansible-deployment/blob/master/playbooks/roles/lxc_hosts/defaults/main.yml#L16-L31

[Kamil Szczygiel (kamil-szczygiel)]

I've encountered this issue - dnsmasq was not handling addresses for containers since there were entries for these containers in /etc/hosts. This led to lack of Internet connectivity inside of the containers. After adding --no-hosts to lxc-system-manage as Kevin suggested, issue is no longer present.