Open Stack

Well we definitely need a better way to get multiple IP addresses onto one
host. The current steps are terrible for a user and even for an
orchestration system like heat. I can't imagine how convoluted a template
would look to automate that process...

I'm not suggesting multiple NICs is the only approach, but I don't think
STP is a very strong excuse. First, if we trust the spoofing filtering of
security groups, looped traffic won't make it out of the other side of the
instance because it won't have the correct MAC on egress. Second, if a
Neutron implementation has no STP protection now, a tenant can just use two
instances with two NICs, bridge on both, and take down both networks (see
diagram).

------------NET 1-------------
   |                          |
====                    ====
| A  |                    | B  |
====                    ====
   |                          |
------------NET 2-------------


On Thu, Apr 17, 2014 at 12:06 AM, Aaron Rosen <aaronorosen at gmail.com> wrote:

> Nova currently is preventing one from attaching multiple nics on the same
> L2. That said I don't think we've clearly determined a use case for having
> multiple nics on the same L2. One reason why we don't allow this is doing
> so would allow a tenant to easily loop the network and cause a bcast storm
> and neutron doesn't have any mechanism today to break these loops today.
> One could just enable STP on ovs to do so though I think we should come up
> with a good use case before allowing this type of thing.
>
>
> On Wed, Apr 16, 2014 at 11:53 PM, Kevin Benton <blak111 at gmail.com> wrote:
>
>> This seems painful for a tenant workflow to get multiple addresses. I
>> would like to improve this during the Juno cycle. What is the limitation
>> that is blocking the multi-nic use cases? Is it Nova?
>>
>>
>> On Wed, Apr 16, 2014 at 11:27 PM, Aaron Rosen <aaronorosen at gmail.com>wrote:
>>
>>> Hi Kevin,
>>>
>>> You'd would just create ports that aren't attached to instances and
>>> steal their ip_addresses from those ports and put those in the
>>> allowed-address-pairs on a port OR you could change the allocation range on
>>> the subnet to ensure these ips were never handed out. That's probably the
>>> right approach.
>>>
>>> Aaron
>>>
>>>
>>> On Wed, Apr 16, 2014 at 10:03 PM, Kevin Benton <blak111 at gmail.com>wrote:
>>>
>>>> Yeah, I was aware of allowed address pairs, but that doesn't help with
>>>> the IP allocation part.
>>>>
>>>> Is this the tenant workflow for this use case?
>>>>
>>>> 1. Create an instance.
>>>> 2. Wait to see what which subnet it gets an allocation from.
>>>> 3. Pick an IP from that subnet that doesn't currently appear to be in
>>>> use.
>>>> 4. Use the neutron-cli or API to update the port object with the extra
>>>> IP.
>>>> 5. Hope that Neutron will never allocate that IP address for something
>>>> else.
>>>>
>>>>
>>>> On Wed, Apr 16, 2014 at 9:46 PM, Aaron Rosen <aaronorosen at gmail.com>wrote:
>>>>
>>>>> Whoops Akihiro beat me to it :)
>>>>>
>>>>>
>>>>> On Wed, Apr 16, 2014 at 9:46 PM, Aaron Rosen <aaronorosen at gmail.com>wrote:
>>>>>
>>>>>> The allowed-address-pair extension that was added here (
>>>>>> https://review.openstack.org/#/c/38230/) allows us to add arbitrary
>>>>>> ips to an interface to allow them. This is useful if you want to run
>>>>>> something like VRRP between two instances.
>>>>>>
>>>>>>
>>>>>> On Wed, Apr 16, 2014 at 9:39 PM, Kevin Benton <blak111 at gmail.com>wrote:
>>>>>>
>>>>>>> I was under the impression that the security group rules blocked
>>>>>>> addresses not assigned by neutron[1].
>>>>>>>
>>>>>>> 1.
>>>>>>> https://github.com/openstack/neutron/blob/master/neutron/agent/linux/iptables_firewall.py#L188
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Apr 16, 2014 at 9:20 PM, Aaron Rosen <aaronorosen at gmail.com>wrote:
>>>>>>>
>>>>>>>> You can do it with ip aliasing and use one interface:
>>>>>>>>
>>>>>>>> ifconfig eth0 10.0.0.22/24
>>>>>>>> ifconfig eth0:1 10.0.0.23/24
>>>>>>>> ifconfig eth0:2 10.0.0.24/24
>>>>>>>>
>>>>>>>> 2: eth0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc mq
>>>>>>>> state DOWN qlen 1000
>>>>>>>>     link/ether 40:6c:8f:1a:a9:31 brd ff:ff:ff:ff:ff:ff
>>>>>>>>     inet 10.0.0.22/24 brd 10.0.0.255 scope global eth0
>>>>>>>>        valid_lft forever preferred_lft forever
>>>>>>>>     inet 10.0.0.23/24 brd 10.0.0.255 scope global secondary eth0:1
>>>>>>>>        valid_lft forever preferred_lft forever
>>>>>>>>     inet 10.0.0.24/24 brd 10.0.0.255 scope global secondary eth0:2
>>>>>>>>        valid_lft forever preferred_lft forever
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On Wed, Apr 16, 2014 at 8:53 PM, Kevin Benton <blak111 at gmail.com>wrote:
>>>>>>>>
>>>>>>>>> Web server running multiple SSL sites that wants to be compatible
>>>>>>>>> with clients that don't support the SNI extension. There is no way for a
>>>>>>>>> server to get multiple IP addresses on the same interface is there?
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Wed, Apr 16, 2014 at 5:50 PM, Aaron Rosen <
>>>>>>>>> aaronorosen at gmail.com> wrote:
>>>>>>>>>
>>>>>>>>>> This is true. Several people have asked this same question over
>>>>>>>>>> the years though I've yet to hear a use case why one really need to do
>>>>>>>>>> this. Do you have one?
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> On Wed, Apr 16, 2014 at 3:12 PM, Ronak Shah <
>>>>>>>>>> ronak at nuagenetworks.net> wrote:
>>>>>>>>>>
>>>>>>>>>>> Hi Vikash,
>>>>>>>>>>> Currently this is not supported. the NIC not only needs to be in
>>>>>>>>>>> different subnet, they have to be in different network as well (container
>>>>>>>>>>> for the subnet)
>>>>>>>>>>>
>>>>>>>>>>> Thanks
>>>>>>>>>>> Ronak
>>>>>>>>>>>
>>>>>>>>>>> On Wed, Apr 16, 2014 at 3:51 AM, Vikash Kumar <
>>>>>>>>>>> vikash.kumar at oneconvergence.com> wrote:
>>>>>>>>>>>
>>>>>>>>>>>> *With 'interfaces' I mean 'nics' of VM*.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> On Wed, Apr 16, 2014 at 4:18 PM, Vikash Kumar <
>>>>>>>>>>>> vikash.kumar at oneconvergence.com> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>>      I want to launch one VM which will have two Ethernet
>>>>>>>>>>>>> interfaces with IP of single subnet. Is this supported now in openstack ?
>>>>>>>>>>>>> Any suggestion ?
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanx
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>> OpenStack-dev mailing list
>>>>>>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>>>>>>>
>>>>>>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> _______________________________________________
>>>>>>>>>>> OpenStack-dev mailing list
>>>>>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> _______________________________________________
>>>>>>>>>> OpenStack-dev mailing list
>>>>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> Kevin Benton
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> OpenStack-dev mailing list
>>>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> OpenStack-dev mailing list
>>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Kevin Benton
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> OpenStack-dev mailing list
>>>>>>> OpenStack-dev at lists.openstack.org
>>>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> OpenStack-dev mailing list
>>>>> OpenStack-dev at lists.openstack.org
>>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Kevin Benton
>>>>
>>>> _______________________________________________
>>>> OpenStack-dev mailing list
>>>> OpenStack-dev at lists.openstack.org
>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>
>>>>
>>>
>>> _______________________________________________
>>> OpenStack-dev mailing list
>>> OpenStack-dev at lists.openstack.org
>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>
>>>
>>
>>
>> --
>> Kevin Benton
>>
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>


-- 
Kevin Benton
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140417/53633266/attachment.html>