cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Imran Ahmed" <im...@eaxiom.net>
Subject RE: OVS Plugin
Date Sat, 15 Apr 2017 01:36:13 GMT
Hi Simon,

That’s very interesting. Do you use openvSwitch  underneath on the
hypervisor or default Linux bridging ? Also what is your preference?

Thanks,

Imran 

-----Original Message-----
From: Simon Weller [mailto:sweller@ena.com] 
Sent: Saturday, April 15, 2017 1:35 AM
To: users@cloudstack.apache.org
Subject: Re: OVS Plugin

We trunk a vlan down to the hosts, make it the KVM traffic label for the
guest network (we run VPCs in advanced mode) and allow ACS to build the
VXLAN VNI within the vlan.

The native VXLAN support in the linux kernel uses multicast, so you do need
an ip on the interface (we use a different network per cluster).


We utilize both Whitebox and Brocade switches, all 10G SFP+ for frontend and
backend connectivity. Each host blade has 4 x 10GB, 2 for the front end and
2 for the storage connectivity.

The VR terminates the guest network, and you can then plugin a vlan in for
public and private gateway access.


We'll actually be presenting our design in Miami at the Cloudstack
Conference, so ilya, you have an excuse to make the trip now!


- Si


________________________________
From: ilya <ilya.mailing.lists@gmail.com>
Sent: Friday, April 14, 2017 3:17 PM
To: users@cloudstack.apache.org
Subject: Re: OVS Plugin

Hi Simon

Would you mind expanding a little more on your setup?

Specifically what is being used underneath.

thanks
ilya

On 4/14/17 9:11 AM, Simon Weller wrote:
> I'd strongly suggest you consider using the native VXLAN support for KVM.
It works extremely well and we run it in production.
>
>
> - Si
>
>
>
>
> ________________________________
> From: Dag Sonstebo <Dag.Sonstebo@shapeblue.com>
> Sent: Friday, April 14, 2017 10:57 AM
> To: users@cloudstack.apache.org
> Subject: Re: OVS Plugin
>
> Hi Imran,
>
> OVS is the same as GRE tunnelling, which you will have as an isolation
method for guest networking – see
http://docs.cloudstack.apache.org/en/latest/networking/ovs-plugin.html.
The OVS Plugin — Apache CloudStack 4.8.0
documentation<http://docs.cloudstack.apache.org/en/latest/networking/ovs-plu
gin.html>
docs.cloudstack.apache.org
Introduction to the OVS Plugin¶ The OVS plugin is the native SDN
implementations in CloudStack, using GRE isolation method. The plugin can be
used by CloudStack to ...



>
> Please let us know how you get on – especially how your hypervisor nodes
cope with CPU load once the GRE tunnels start growing in numbers
(historically this has not scaled well).
>
> Regards,
> Dag Sonstebo
> Cloud Architect
> ShapeBlue
>
> On 13/04/2017, 21:37, "Imran Ahmed" <imran@eaxiom.net> wrote:
>
>     Dear Team,
>
>     I have setup cloudstack 4.9 with KVM hypervisor and advanced
networking on
>     CentOS7. Also we installed and setup openvswitch on the hypervisors
(KVM)
>     hosts.
>
>     Below are traffic labels for kvm
>
>     Cloudbr0  for management
>     Cloudbr1 for guest
>     Cloudbr2 for public
>
>     After configuring the zone, pod , cluster, host, primary and secondary
>     storages we wanted to enable the OVS plugin under service providers
for the
>     guest network.
>
>     However OVS is not shown in the list.
>
>     Please suggest what could be wrong here.
>
>     Kind regards,
>
>     Imran
>
>
>
>
> Dag.Sonstebo@shapeblue.com
> www.shapeblue.com<http://www.shapeblue.com>
Shapeblue - The CloudStack Company<http://www.shapeblue.com/>
www.shapeblue.com
Background Cloudstack relies on a fixed download site when it fetches the
built-in guest VM templates. That download site has historically



> 53 Chandos Place, Covent Garden, London  WC2N 4HSUK
> @shapeblue
>
>
>
>


Mime
View raw message