[openstack-dev] [Neutron][L3] BGP Dynamic Routing : Change in openstack/neutron-specs[master]: BGP dynamic routing
keshava.a at hp.com
Mon Jun 30 04:09:06 UTC 2014
Please share any blue-print, information about how dr-agent talks to BGP routing protocol running in Provider router ?
That means dr-agent also running BGP routing protocol ?
Please share that design/blue-print ?
Thanks & Regards,
From: loy wolfe [mailto:loywolfe at gmail.com]
Sent: Saturday, June 28, 2014 8:23 AM
To: OpenStack Development Mailing List (not for usage questions)
Subject: Re: [openstack-dev] [Neutron][L3] BGP Dynamic Routing : Change in openstack/neutron-specs[master]: BGP dynamic routing
On Sat, Jun 28, 2014 at 3:57 AM, A, Keshava <keshava.a at hp.com<mailto:keshava.a at hp.com>> wrote:
I have added the + community in this List.
Please find my thoughts how it may/can evolve later once BGP is introduced.
Data center will host multi-Tennant with overlapping address amount tenants with private IP address.
When these Data centers are geographically separated, and connected over Service provider network they will continue to communicate to each other over there Private IP address only.
Since there is overlapping , these “ private-IP needs to be injected with corresponding VPN into BGP respective vrf-table”
We may have multiple options
1. Learn these IP dynamically by each BGP inject this IP to Gate way BGP.
2. Inject all the VM ip to Gateway router by Controller.
==> OpenStack Controller to add the private Prefix ?
==> Or ODL controller to add ?
because VM state is first known by nova, then to Neutron by vif binding . The neutron server send these message to dr-agent, dr-agent speak BGP to provider router. No need for ODL, either no need for any CN/NN and data-plane gateway to speak BGP.
When VM moves dynamically to different subnet/Rack: : adding by controller and then advertising this new prefix may be too late ?
there is callback notify for sync between nova and neutron, since Icehouse. however, if you really need this kind of strict order, you have to change nova compute driver to pend, waiting for neutron finished it's work.
Let me know the opinions.
[cid:image001.png at 01CF9447.1D7A7FF0]
Thank & regards,
From: Carl Baldwin (Code Review) [mailto:review at openstack.org]
Sent: Friday, June 27, 2014 9:51 AM
To: Artem Dmytrenko; Jaume Devesa
Cc: mark mcclain; Sean M. Collins; Anthony Veiga; Pedro Marques; Nachi Ueno; YAMAMOTO Takashi; Itsuro Oda; fumihiko kakuma; A, Keshava
Subject: Change in openstack/neutron-specs[master]: BGP dynamic routing
Carl Baldwin has posted comments on this change.
Change subject: BGP dynamic routing
Patch Set 8: Code-Review-1
Still concerned about the first diagram. Ping me tomorrow if you want.
Line 29: Allow Neutron to dynamically announce to from external uplink
nit: maybe insert "and/or" here between to and from.
Line 50: gateway_router[color = red];
In this topology, the red router is still the single router between the Openstack network and the uplinks. So, there are not multiples paths. I think openstack would have to peer with the gateway and then the gateway would have to peer with the uplinks because it is the next hop for both uplinks.
The red gateway router needs to be configured with the neutron networks anyway so why would it need to learn them from openstack through BGP? Why wouldn't the gateway router simply peer with the uplinks and announce what it knows? I don't see a need for BGP in openstack with this diagram.
Line 93: uplink_router1;
This diagram is more like what I was thinking. Here, it might make sense for openstack to speak with the uplinks.
To view, visit https://review.openstack.org/90833
To unsubscribe, visit https://review.openstack.org/settings
Gerrit-Owner: Artem Dmytrenko <nextone92 at yahoo.com<mailto:nextone92 at yahoo.com>>
Gerrit-Reviewer: Anthony Veiga <anthony_veiga at cable.comcast.com<mailto:anthony_veiga at cable.comcast.com>>
Gerrit-Reviewer: Artem Dmytrenko <nextone92 at yahoo.com<mailto:nextone92 at yahoo.com>>
Gerrit-Reviewer: Carl Baldwin <carl.baldwin at hp.com<mailto:carl.baldwin at hp.com>>
Gerrit-Reviewer: Itsuro Oda <oda at valinux.co.jp<mailto:oda at valinux.co.jp>>
Gerrit-Reviewer: Jaume Devesa <devvesa at gmail.com<mailto:devvesa at gmail.com>>
Gerrit-Reviewer: Nachi Ueno <nachi at ntti3.com<mailto:nachi at ntti3.com>>
Gerrit-Reviewer: Pedro Marques <pedro.r.marques at gmail.com<mailto:pedro.r.marques at gmail.com>>
Gerrit-Reviewer: Sean M. Collins <sean_collins2 at cable.comcast.com<mailto:sean_collins2 at cable.comcast.com>>
Gerrit-Reviewer: YAMAMOTO Takashi <yamamoto at valinux.co.jp<mailto:yamamoto at valinux.co.jp>>
Gerrit-Reviewer: fumihiko kakuma <kakuma at valinux.co.jp<mailto:kakuma at valinux.co.jp>>
Gerrit-Reviewer: keshava <keshava.a at hp.com<mailto:keshava.a at hp.com>>
Gerrit-Reviewer: mark mcclain <mmcclain at yahoo-inc.com<mailto:mmcclain at yahoo-inc.com>>
OpenStack-dev mailing list
OpenStack-dev at lists.openstack.org<mailto:OpenStack-dev at lists.openstack.org>
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 54204 bytes
More information about the OpenStack-dev