[Openstack] [neutron] GRE network with 5 nodes

Ruzicka, Marek marek.ruzicka at t-systems.sk
Wed May 14 09:45:54 UTC 2014

We used to have similar issues with GRE and packet loss/latency in Grizzly. It was caused by wrong (low) MTU size settings on ‘GRE interfaces’. I believe 1546 is the lowest working value (it is part of the official docu these days, it was not back then) Ever since we moved to jumbo frames, all the troubles are gone.

From: zengshan2008 [mailto:zengshan2008 at gmail.com]
Sent: 14. mája 2014 11:15
To: Ageeleshwar Kandavelu; gustavo panizzo <gfa>; <openstack at lists.openstack.org>
Subject: Re: [Openstack] [neutron] GRE network with 5 nodes

Did you found the same problem when using Grizzly version?

发件人:Ageeleshwar Kandavelu <Ageeleshwar.Kandavelu at csscorp.com>
发送时间:2014-05-14 17:00
主题:Re: [Openstack] [neutron] GRE network with 5 nodes
收件人:"gustavo panizzo <gfa>"<gfa at zumbi.com.ar>,"<openstack at lists.openstack.org>"<openstack at lists.openstack.org>

I wasn't using ml2. I did a havana with neutron-openvswitch-agent. the old way. Its not only the latency. When I created external network and assigned floatingip and tried to reach it only one or two packets where reaching. it was 99% packet loss.
From: gustavo panizzo <gfa> [gfa at zumbi.com.ar]
Sent: Wednesday, May 14, 2014 2:24 PM
To: Ageeleshwar Kandavelu; <openstack at lists.openstack.org>
Subject: Re: [Openstack] [neutron] GRE network with 5 nodes

On 05/14/2014 03:56 AM, Ageeleshwar Kandavelu wrote:
> What I inferred was while using the GRE mode the switch's port were
> constantly blinking indicating heavy traffic. I am aware that in GRE
> mode there is a mesh of tunnels between the various nodes. I assumed
> that the GRE mode was placing a burden too heavy on the switch and tore
> down the setup and created a vlan setup instead. The performance was
> good with vlan.
> Is this expected or I had been doing something wrong?
i've observed the same behavior, gre adds more latency than vlan

were you creating new VMs? ml2 pop layer plugin should help in that case

1AE0 322E B8F7 4717 BDEA BF1D 44BB 1BA7 9F6C 6333

Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack at lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140514/050092e8/attachment.html>

More information about the Openstack mailing list