On Fri, Apr 11, 2014 at 12:23 PM, Brian Candler <[email protected]> wrote:
> On 11/04/2014 18:00, Gurucharan Shetty wrote:
>>
>> Open vSwitch starts a little later than networking, I think. I would
>> expect
>> a race condition where the bridges are not yet created and there is an
>> attempt to add an IP address.
>>
>> I haven't looked at 14.04 to see if such a race condition cannot
>> exist. If such a race condition does not exist, it may be a nice
>> strategy.
>
> From what I gather, it was a known problem before, but seems to have been
> fixed:
> https://bugs.launchpad.net/ubuntu/+source/openvswitch/+bug/1084028
>
> There is now some upstart magic. I see /etc/init/openvswitch-switch.conf:
>
> # vim: set ft=upstart ts=2 et:
> description "Open vSwitch switch"
> author "James Page <[email protected]"
>
> start on (local-filesystems and net-device-up IFACE=lo)
> stop on runlevel [!2345]
> ... etc, eventually calls ovs_ctl start --system-id=random

Thanks for looking this up. At first glance, it looks like there will
be a problem if /usr is loaded through NFS. (All OVS utilities are
placed in /usr).
I guess, if there are no bug reports so far, no one really keeps their
/usr on NFS.

>
>
>>
>>> You don't need to do that with bridge-utils. You'd simply write:
>>>
>>> auto eth0
>>> iface eth0 inet manual
>>>
>>> auto br-wan
>>> iface br-wan inet dhcp
>>>      bridge_ports eth0
>>
>> In this case, someone would already need to create the bridges. You
>> are only using this to assign IP addresses, right?
>
> No: what I've shown by itself is sufficient to create the bridge and add the
> port(s) into it when the system starts, and also assign an IP address to the
> bridge.
Okay.

>
> I realise openvswitch has an extra level of indirection (ports can contain
> one or more interfaces, whereas in brctl you connect interfaces directly to
> the bridges). But it seems non-DRY to say bridge X contains ports Y Z, and
> also that the ports Y Z belong to bridge X.
I agree, that is not nice. There were a couple of underlying reasons
for that approach.
For one, when I wrote that up, I tested in Ubuntu 11.04 and it had a
bug/constraint in the ifupdown package that forced me to do it.



>
> Regards,
>
> Brian.
>
_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss

Reply via email to