On Mon, May 21, 2012 at 06:44:16PM +0200, Jiri Pirko wrote:
Mon, May 21, 2012 at 01:54:26PM CEST, rpazdera(a)redhat.com wrote:
>I have another idea for a feature to LNST. I'm using it together with
>libvirt and I
>thought that it could be nice to have direct control over the network
>from LNST. When using libvirt for testing, this could be easily
>the virsh command.
>Now, when you're running some tests, you have to set up the required network
>topology by yourself and put MAC and IP addresses into the recipe. When you
>want to run the test on a different VM (with different version of
>you need to set up the topology again by hand and also alter the recipe.
>I thought we could make LNST to be able to control this setup.
>Virtually all the
>configuration of libvirt's networks and domains are accessible through XML,
>so they can be easily parsed and modified.
>There could be a pool of installed VM's available on the controller.
>LNST could detect
>available machines (virsh list --all), modify the number of
>interfaces they have
>(virsh dumpxml/define), define some networks (bridges/tap devices -
>and connect them into some topology. Everything would be controlled
>the recipe XML file. At the end LNST would boot up the machines and
>run the tests.
>Some of those machines could act as switches (with openvswitch) or
>even routers so
>we could build a whole virtual network.
>What do you think?
>Could this approach be useful?
I was thinking about something like this a while ago. Your approach
looks very interesting to me. I would suggest if we meet Probably with
Jan and talk about his some more.
But what I think needs to be done first it to propose xml elements for
this, after that to implement stuff.
That's probably a good idea. I'll try to outline how this could work
in practice first, think about the XML elements more and then we can
decide wether it's worth it and maybe meet up and with Honza and
discuss it further.
>LNST-developers mailing list