Two Internal Devices Found in a Router Namespace

[Stephen Ma]

Doing a sudo ip netns qrouter-<router-id> ifconfig for a router namespace reveals the following. There are two internal devices (qr-XXXX) in this namespace. When a VM is created with a floating ip using this router, it cannot be pinged.

lo Link encap:Local Loopback
          inet addr:127.0.0.1 Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING MTU:16436 Metric:1
          RX packets:147 errors:0 dropped:0 overruns:0 frame:0
          TX packets:147 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:12936 (12.9 KB) TX bytes:12936 (12.9 KB)

qg-c9bad8cf-63 Link encap:Ethernet HWaddr fa:16:3e:fe:20:2d
          inet addr:15.184.41.158 Bcast:15.184.47.255 Mask:255.255.248.0
          inet6 addr: fe80::f816:3eff:fefe:202d/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
          RX packets:7567222 errors:0 dropped:235502 overruns:0 frame:0
          TX packets:2831290 errors:0 dropped:4 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:14540514463 (14.5 GB) TX bytes:246815068 (246.8 MB)

qr-57bb264d-7c Link encap:Ethernet HWaddr fa:16:3e:8f:dc:be
          inet addr:10.24.0.1 Bcast:10.24.255.255 Mask:255.255.0.0
          inet6 addr: fe80::f816:3eff:fe8f:dcbe/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
          RX packets:292 errors:0 dropped:0 overruns:0 frame:0
          TX packets:3 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:23856 (23.8 KB) TX bytes:126 (126.0 B)

qr-9883db4f-07 Link encap:Ethernet HWaddr fa:16:3e:5e:70:09
          inet addr:10.24.0.1 Bcast:10.24.255.255 Mask:255.255.0.0
          inet6 addr: fe80::f816:3eff:fe5e:7009/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
          RX packets:171988 errors:0 dropped:0 overruns:0 frame:0
          TX packets:319691 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:14116358 (14.1 MB) TX bytes:653426894 (653.4 MB)

[Stephen Ma (stephen-ma)]

The question is why there are 2 internal devices with the same IP addresses. In this case, what happened is that the user issued the command "neutron router-interface-delete <router-id> <subnet-id>". The API-server processed the request and then it issued the routers updated notification. However while the the notification is in flight, the machine running the L3-agent hosting the router crashed. After reboot, L3-agent started up again, but it "forgot" about needing to delete the internal device from the namespace.

After awhile, the user created ran the command "neutron router-interface-add <router-id> <subnet-id>". Now the L3-agent created another internal device in the namespace.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/57954

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/57954
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=b0306a93645ea6475a2bc045d8fc8bc4bd6f00a5
Submitter: Jenkins
Branch: master

commit b0306a93645ea6475a2bc045d8fc8bc4bd6f00a5
Author: Stephen Ma <email address hidden>
Date: Wed Nov 20 14:52:23 2013 -0800

    Delete duplicate internal devices in router namespace

    When neutron router-interface-delete <router> <subnet> is ran
    during L3-agent restart, the agent may fail to delete the old
    internal device. After the restart, when the command "neutron
    router-interface-add <router> <subnet>" is ran again, the
    router ends up having two internal devices configured with the
    same IP address.

    Closes-Bug: #1244853
    Change-Id: I0d7e2db6aa7dae26d0fc3fe1b1527762cb1e3b23