[ovs-discuss] Need to reset controller after LXC interface goes (back) up?
Bertrand Bonnefoy-Claudet
bertrandbc at gmail.com
Fri Oct 25 17:11:10 UTC 2013
Hi,
I'm trying to connect two LXC hosts, h00 and h10 with Open vSwitch (one bridge)
on Debian (testing) and I can't have them communicate unless I reset the
controller setting of the bridge.
Here's what I have configured (you can ignore br1):
root at debtest# ovs-vsctl show
31135ada-b505-42dc-bb8d-6fad98553d88
Bridge "br1"
Controller "ptcp:"
Port "br1"
Interface "br1"
type: internal
Port "eth10"
Interface "eth10"
Port "eth11"
Interface "eth11"
Bridge "br0"
Controller "ptcp:"
Port "br0"
Interface "br0"
type: internal
Port "eth00"
Interface "eth00"
Port "eth01"
Interface "eth01"
ovs_version: "1.4.2"
Host h00 connects to eth00 (br0) and h10 connects to eth01 (br0). I give them
two IP addresses on the same subnet and when the controller is set to ptcp:,
they can ping each other. That's nice already.
controller
|
|
+--+---------+
| | |
| br0 | LXC
| | |
| +--eth00--+----h00
| | |
| +--eth01--+----h10
| |
+------------+
Here's my problem. If I reboot one of my LXC hosts and then give it back its IP
address, there is no connectivity anymore. Reissuing the following command
solves the problem temporarily but that's not satisfying:
ovs-vsctl set-controller br0 ptcp:
The same applies if some host joins the network via a another port. Any
idea why such a behaviour is encountered?
LXC configuration of one of the hosts:
lxc.utsname = h00
lxc.network.type = veth
lxc.network.flags = up
lxc.network.veth.pair = eth00
lxc.network.hwaddr = 02:02:02:02:02:00
lxc.rootfs = /var/lib/lxc/h00/rootfs
lxc.tty = 4
lxc.pts = 1024
--
Bertrand
More information about the discuss
mailing list