[Openstack] subnet gateway's arp ack not sent back

Liu Wenmao marvelliu at gmail.com
Wed Apr 17 11:04:34 UTC 2013


Hi all:

I set up openstack with quantum successfully, but I use floodlight as the
network controller, VMs can not ping their gateway.

I use a host as compute/network controller(30.0.0.1), and another host as a
compute node(30.0.0.11). The VM X address is 100.0.0.7 and the subnet
gateway G is 100.0.0.1. I use namespace to isolate networks (floodlight
restproxy seems not to support namespace, but I use floodligt standalone)

When X is pinging G, I can see gateway responses a ARP ack:

root at controller:/usr/src/floodlight# ip netns exec
qrouter-7bde1209-e8ed-4ae6-a627-efaa148c743c tcpdump -nn -i qr-8af2e01f-bb
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on qr-8af2e01f-bb, link-type EN10MB (Ethernet), capture size
65535 bytes
18:52:32.769334 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:52:32.769371 ARP, Reply 100.0.0.1 is-at fa:16:3e:f7:3d:5e, length 28
18:52:33.769049 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:52:33.769082 ARP, Reply 100.0.0.1 is-at fa:16:3e:f7:3d:5e, length 28
18:52:34.769117 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:52:34.769149 ARP, Reply 100.0.0.1 is-at fa:16:3e:f7:3d:5e, length 28

But when I listen to the bridge br-int or physical interface eth2, no ARP
reply is heard:

root at controller:/usr/src/floodlight# tcpdump -i br-int -nn
tcpdump: WARNING: br-int: no IPv4 address assigned
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on br-int, link-type EN10MB (Ethernet), capture size 65535 bytes
18:50:31.405691 IP 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request
from fa:16:3e:1c:65:d0, length 286
18:50:31.749137 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:50:32.749232 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:50:33.749575 ARP, Request who-has 100.0.0.1 tell 100.0.0.7, length 28



root at controller:/usr/src/floodlight# tcpdump -i eth2 proto gre -nn
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth2, link-type EN10MB (Ethernet), capture size 65535 bytes
18:54:28.784500 IP 30.0.0.11 > 30.0.0.1: GREv0, key=0x0, length 50: ARP,
Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:54:29.784430 IP 30.0.0.11 > 30.0.0.1: GREv0, key=0x0, length 50: ARP,
Request who-has 100.0.0.1 tell 100.0.0.7, length 28
18:54:30.784317 IP 30.0.0.11 > 30.0.0.1: GREv0, key=0x0, length 50: ARP,
Request who-has 100.0.0.1 tell 100.0.0.7, length 28

After I delete the controller from openvswitch and restart openvswithes,
VMs can ping their gateway. I do not what causes the problem

Can anyone give me some resources how the namespace and bridges work
together.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20130417/a184875b/attachment.html>


More information about the Openstack mailing list