Open Stack

Dear Joe

Thank you very much for the reply!

The answer is very helpful for me to understand what multi-host mode of
nova-network exactly provides. By the way, in aspect of reliability,
robustness and fault-tolerance for networking service on OpenStack,
nova-network still seems better than neutron, the neutron provides lots of
useful networking features though. I'd like to hear comments about this.

I'm now mainly focusing on analyzing how much high-availability can be
guaranteed by each networking service type. I know the latest neutron
provides distributed L3 service, i.e., DVR (Distributed Virtual Router) and
can give high-availability via Pacemaker or something like that, but I'm
not sure this can be said that the neutron is obviously better than
nova-network in terms of the service continuity.

Regards

Byeong-Gi

2015-05-08 0:56 GMT+09:00 Joe Topjian <joe at topjian.net>:

> Hello,
>
> If the nova-network service is down, then only actions that would involve
> nova-network (creating and terminating instances for example) won't work.
> Instances that are already running will still be able to communicate with
> both the outside network and other instances in the cloud.
>
> You can easily test this by just stopping the nova-network service on the
> compute node (assuming you have a window where no one will be launching
> instances).
>
> multi-host provides continuity in that each compute node becomes a network
> gateway for the instances hosted on that node. If that compute node is
> physically down, instances on all other compute nodes can still have
> external network access.
>
> Contrast this with single-host networking where all traffic is routed
> through the single host. (Again, nova-network does not need to be running
> on that single host for traffic to still get out.)  But if that single host
> is physically down, then *no* instances in your cloud have external network
> access.
>
> Hope that helps,
> Joe
>
> On Thu, May 7, 2015 at 1:08 AM, BYEONG-GI KIM <kimbyeonggi at gmail.com>
> wrote:
>
>> Hello.
>>
>> It seems that this question would be quite outdated question, because
>> this is a question about nova-network instead of neutron.
>>
>> I wonder whether VMs located in a Compute Node, e.g., Compute A, are
>> accessible while its nova-network service is down if the other nova-network
>> is running on the other Compute Nodes, such as Compute B, Compute C, etc.
>>
>> Or, does the multi-host just provide continuity of the networking service
>> via avoiding single point failure?
>>
>> Thanks in advance!
>>
>> Regards,
>> Byeong-gi
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20150508/725e4ade/attachment.html>