Ask Your Question

bennybrit's profile - activity

2016-10-27 14:16:47 -0500 received badge  Famous Question (source)
2016-07-06 11:11:54 -0500 received badge  Notable Question (source)
2016-06-09 23:01:15 -0500 received badge  Popular Question (source)
2016-06-06 12:58:15 -0500 asked a question DHCPOFFER getting dropped at tap device

Hi,

I have an issue that i'm stuck at in the last week, i have a 3 node Kilo environment (1 Controller, 1 Network, 1 Compute). I'm running an instance with two network interfaces (the first is connected to private tenant network, and the second connected to a flat provider network), the first interface gets DHCP without any problems, the second one not.

I have run tcpdump all the way from the network node till the instance: i see DHCPDiscover gets all the way from the instance to the dnsmasq proccess, and i see DHCPOffer goes all the way from the DNSMasq to the compute node eth4(used for provider network) => phy-br-ex => int-br-ex => qvo interface => qvb interface => qbr bridge and here it's lost (don't see it at the tap interface).

iptables and ebtales are empty and have no rules.

below is the ml2 configuration from the compute node:

[ml2]
type_drivers = flat,vlan,gre,vxlan
tenant_network_types = gre
mechanism_drivers = openvswitch

[ml2_type_flat]
flat_networks = external

[ml2_type_gre]
tunnel_id_ranges = 1:1000

[securitygroup]
enable_security_group = True
enable_ipset = True
firewall_driver = neutron.agent.firewall.NoopFirewallDriver

[ovs]
local_ip = 192.168.1.102
bridge_mappings = external:br-ex

[agent]
tunnel_types = gre

ml2 configuration from the network node:

[ml2]
type_drivers = flat,vlan,gre,vxlan
tenant_network_types = gre
mechanism_drivers = openvswitch

[ml2_type_flat]
flat_networks = external

[ml2_type_gre]
tunnel_id_ranges = 1:1000

[securitygroup]
enable_security_group = True
enable_ipset = True
firewall_driver = neutron.agent.firewall.NoopFirewallDriver

[ovs]
local_ip = 192.168.1.101
bridge_mappings = external:br-ex

[agent]
tunnel_types = gre

My analysis so far:

  • It's not the OVS flow (that's why i'm not posting them here), i tripled checked them and all looks fine, and as i wrote the DHCPOffer arrives to the qbr bridge (which is LinuxBridge)

  • It's not the iptables/ebtables, all the tables are empty, and as you can see from my config i'm using: neutron.agent.firewall.NoopFirewallDriver

  • When configuring the ip address manually on the instance, it goes "crazy" - looses connection, sometimes answer to ping sometimes not

I'm suspecting it's something in the qbr bridge or the tap device but can't figure it out, and already started to pull my hair off :-)