Open Stack

Hey Ned!

Thoughts in line...





On 2/17/16, 12:02 PM, "Ned Rhudy (BLOOMBERG/ 731 LEX)" <erhudy at bloomberg.net> wrote:

>
>We're in mostly the same boat; using nova-network with VLAN segmentation and looking at a Neutron migration (though ours may take a more drastic path and take us to Neutron+Calico). One question I have for you: the largest issue and conceptual leap we had when
> initially prototyping Neutron+linuxbridge was that our current model only has controllers and work nodes, with no provisions for dedicated network nodes to route in/out of the cluster. All our work nodes can route by themselves, which would have steered us
> towards a DVR model, but that seems to have its own issues as well as mandating OVS.
>
>
>Since your branch indicates you're using linuxbridge on Icehouse, are you provisioning network nodes as part of your migration, or are you avoiding needing to provision network nodes in a different fashion?

We're in a similar situation... We have control nodes and hypervisors (work nodes). We put the nova-network L3 functionality on the control nodes (so not a DVR/"nova-networking-ha" model). I'm still working out the specific details of the L3 migration, but in the lab I've been able to create a corresponding logical router, set the upstream gateway, and then assign floating Ips to tenants. My next step was going to be to map them over (like we've done for the L2 ports) such that starting the L3 agent plumbs everything the way it currently is in nova-network, with l3 still homed on the control nodes. Alternatively, I've created a standalone hardware router, plumbed the network into the existing tenant VLAN in a VRF, and handed control over to the corresponding L3 plugin for that hardware (this is why I added functionality to keep the tenant networks on the same VLAN instead of simply assigning a new one).

I also know of some people who are putting the neutron l3 agents on the hypervisors. I don't think they have any special scheduling done to make sure floaters are mapped to Vms on that particular node, but that may work for you as well, and kind of resembles the nova-network-ha model... Especially if you map the floaters to agents on the physical node the floater is currently on you'd save yourself the ARP trouble. I don't want to throw them under the bus, so I'll let them speak up if they care to talk about that :-)

As far as Calico goes, I wouldn't even know where to start... I've not looked at it enough to have any thoughts.

I hope that helps. I'm going to be at the summit one way or another so I'd love to find some time to meet up and chat about it in person if you'd like.

-- Kevin

>
>From: kevinbri at cisco.com 
>Subject: Re: [Openstack-operators] Nova-network -> Neutron Migration
>
>
>
>Definitely, I can work on that. I need to get the migration done first, but once I do I plan to open source our plays and whatever else to help people perform the migration themselves. At that point I can work on adding some stuff to the networking
> guide as well. Probably will be a few months from now, though.
>
>
>
>
>On 2/17/16, 9:29 AM, "Matt Kassawara" <mkassawara at gmail.com> wrote:
>
>>Cool! I'd like to see this stuff in the networking guide... or at least a link to it for now.
>>
>>On Wed, Feb 17, 2016 at 8:14 AM, Kevin Bringard (kevinbri)
>><kevinbri at cisco.com> wrote:
>>
>>Hey All!
>>
>>I wanted to follow up on this. We've managed successfully migrated Icehouse with per tenant networks (non overlapping, obviously) and L3 services from nova-networking to neutron in the lab. I'm working on the automation bits, but once that is done we'll start
>> migrating real workloads.
>>
>>I forked Sam's stuff and modified it to work in icehouse with tenants nets: 
>>https://github.com/kevinbringard/novanet2neutron/tree/icehouse <https://github.com/kevinbringard/novanet2neutron/tree/icehouse>.
> I need to update the README to succinctly reflect the steps, but the code is there (I'm going to work on the README today).
>>
>>If this is something folks are interested in I proposed a talk to go over the process and our various use cases in Austin:
>>
>>https://www.openstack.org/summit/austin-2016/vote-for-speakers/presentation/7045 <http://www.openstack.org/summit/austin-2016/vote-for-speakers/presentation/7045> <https://www.openstack.org/summit/austin-2016/vote-for-speakers/presentation/7045>
>>
>>-- Kevin
>>
>>
>>
>>On 12/9/15, 12:49 PM, "Kevin Bringard (kevinbri)" <kevinbri at cisco.com> wrote:
>>
>>>It's worth pointing out, it looks like this only works in Kilo+, as it's written. Sam pointed out earlier that this was what they'd run it on, but I verified it won't work on earlier versions because, specifically, in the migrate-secgroups.py it inserts into
>> the default_security_group table, which was introduced in Kilo.
>>>
>>>I'm working on modifying it. If I manage to get it working properly I'll commit my changes to my fork and send it out.
>>>
>>>-- Kevin
>>>
>>>
>>>
>>>On 12/9/15, 10:00 AM, "Edgar Magana" <edgar.magana at workday.com> wrote:
>>>
>>>>I did not but more advanced could mean a lot of things for Neutron. There are so many possible scenarios that expecting to have a “script” to cover all of them is a whole new project. Not sure we want to explore than. In the past we were recommending to
>>>> make the migration in multiple steps, maybe we could use this as a good step 0.
>>>>
>>>>
>>>>Edgar
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>From: "Kris G. Lindgren"
>>>>Date: Wednesday, December 9, 2015 at 8:57 AM
>>>>To: Edgar Magana, Matt Kassawara, "Kevin Bringard (kevinbri)"
>>>>Cc: OpenStack Operators
>>>>Subject: Re: [Openstack-operators] Nova-network -> Neutron Migration
>>>>
>>>>
>>>>
>>>>Doesn't this script only solve the case of going from flatdhcp networks in nova-network to same dchp/provider networks in neutron. Did anyone test to see if it also works for doing more advanced nova-network configs?
>>>>
>>>>
>>>>___________________________________________________________________
>>>>Kris Lindgren
>>>>Senior Linux Systems Engineer
>>>>GoDaddy
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>From: Edgar Magana <edgar.magana at workday.com>
>>>>Date: Wednesday, December 9, 2015 at 9:54 AM
>>>>To: Matt Kassawara <mkassawara at gmail.com>, "Kevin Bringard (kevinbri)" <kevinbri at cisco.com>
>>>>Cc: OpenStack Operators <openstack-operators at lists.openstack.org>
>>>>Subject: Re: [Openstack-operators] Nova-network -> Neutron Migration
>>>>
>>>>
>>>>
>>>>Yes! We should but with a huge caveat that is not not supported officially by the OpenStack community. At least the author wants to make a move with the Neutron team to make it part of the tree.
>>>>
>>>>
>>>>Edgar
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>From: Matt Kassawara
>>>>Date: Wednesday, December 9, 2015 at 8:52 AM
>>>>To: "Kevin Bringard (kevinbri)"
>>>>Cc: Edgar Magana, Tom Fifield, OpenStack Operators
>>>>Subject: Re: [Openstack-operators] Nova-network -> Neutron Migration
>>>>
>>>>
>>>>
>>>>Anyone think we should make this script a bit more "official" ... perhaps in the networking guide?
>>>>
>>>>On Wed, Dec 9, 2015 at 9:01 AM, Kevin Bringard (kevinbri)
>>>><kevinbri at cisco.com> wrote:
>>>>
>>>>Thanks, Tom, Sam, and Edgar, that's really good info. If nothing else it'll give me a good blueprint for what to look for and where to start.
>>>>
>>>>
>>>>
>>>>On 12/8/15, 10:37 PM, "Edgar Magana" <edgar.magana at workday.com> wrote:
>>>>
>>>>>Awesome code! I just did a small testbed test and it worked nicely!
>>>>>
>>>>>Edgar
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>On 12/8/15, 7:16 PM, "Tom Fifield" <tom at openstack.org> wrote:
>>>>>
>>>>>>On 09/12/15 06:32, Kevin Bringard (kevinbri) wrote:
>>>>>>> Hey fellow oppers!
>>>>>>>
>>>>>>> I was wondering if anyone has any experience doing a migration from nova-network to neutron. We're looking at an in place swap, on an Icehouse deployment. I don't have parallel
>>>>>>>
>>>>>>> I came across a couple of things in my search:
>>>>>>>
>>>>>>>
>>>>https://wiki.openstack.org/wiki/Neutron/MigrationFromNovaNetwork/HowTo <https://wiki.openstack.org/wiki/Neutron/MigrationFromNovaNetwork/HowTo>
>>>>>>>
>>>>http://docs.openstack.org/networking-guide/migration_nova_network_to_neutron.html <http://docs.openstack.org/networking-guide/migration_nova_network_to_neutron.html>
>>>>>>>
>>>>>>> But neither of them have much in the way of details.
>>>>>>>
>>>>>>> Looking to disrupt as little as possible, but of course with something like this there's going to be an interruption.
>>>>>>>
>>>>>>> If anyone has any experience, pointers, or thoughts I'd love to hear about it.
>>>>>>>
>>>>>>> Thanks!
>>>>>>>
>>>>>>> -- Kevin
>>>>>>
>>>>>>NeCTAR used this script (https://github.com/NeCTAR-RC/novanet2neutron )
>>>>>>with success to do a live nova-net to neutron using Juno.
>>>>>>
>>>>>>
>>>>>>
>>>>>>_______________________________________________
>>>>>>OpenStack-operators mailing list
>>>>>>OpenStack-operators at lists.openstack.org
>>>>>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>>>>_______________________________________________
>>>>>OpenStack-operators mailing list
>>>>>OpenStack-operators at lists.openstack.org
>>>>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>>>_______________________________________________
>>>>OpenStack-operators mailing list
>>>>OpenStack-operators at lists.openstack.org
>>>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>_______________________________________________
>>>OpenStack-operators mailing list
>>>OpenStack-operators at lists.openstack.org
>>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>_______________________________________________
>>OpenStack-operators mailing list
>>OpenStack-operators at lists.openstack.org
>>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>
>>
>>
>>
>>
>>
>>
>_______________________________________________
>OpenStack-operators mailing list
>OpenStack-operators at lists.openstack.org
>http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>
>
>
>
>
>