Re: [libvirt] [PATCH v5] ESX: Add routines to interface driver
(Please ignore earlier messages, as for reason the messages got truncated)
Thanks Matthias for the explanation, I was working on the project which required libvirt
to control/set networking on a given ESX host. As ESXi is different from hosted
hypervisors, we need to configure ESX interface (IP, routing, bridge etc), also it is an
integral part of the host settings.
But re-thinking the whole design in light of the explanation provided by you, libvirt
interfaces are actually mapping physical interfaces to virtual network. Further,
hostVirtualSwitch behaves like switch which defines ports where we can plug VM virtual
ethernet cards. It seems likely that we still want to have ability of configuring ESXi
virtual interfaces (hostVirtualNics) so not sure if listing physical Nics is completly
right to get an 100% operational ESX host.
Michal has already reviewed the code and I would do the same if this design is one that
maps to libvirt terminology the most. Though my design is not accpeted, but it was an
amazing learning experience. Thanks for everyones input that came along the way.
Regards,
Ata
Date: Sun, 5 Aug 2012 22:45:26 +0200
Subject: Re: [libvirt] [PATCH v5] ESX: Add routines to interface driver
From: matthias.bolte(a)googlemail.com
To: ata.husain(a)hotmail.com
CC: libvir-list(a)redhat.com
2012/8/5 Matthias Bolte <matthias.bolte(a)googlemail.com>:
> 2012/8/2 Ata E Husain Bohra <ata.husain(a)hotmail.com>:
>> Add following routines to esx_interface_driver:
>> esxNumOfInterfaces,
>> esxNumOfDefinedInterfaces,
>> esxListInterfaces,
>> esxListDefinedInterfaces,
>> esxInterfaceLookupByMACString,
>> esxInterfaceGetXMLDesc,
>> esxInterfaceUndefine,
>> esxInterfaceCreate,
>> esxInterfaceDestroy
>>
>> Signed-off-by: Ata E Husain Bohra <ata.husain(a)hotmail.com>
>
> Okay, I finally had time to take a detailed look at the interface
> driver and the related network driver and your proposed implementation
> for the interface driver.
>
> I think listing HostVirtualNics in the interface driver is not the
> correct mapping between vSphere API and libvirt. Also listing
> HostVirtualNics as bridges is not correct because a HostVirtualNic is
> not a bridge. A HostVirtualSwitch can be seen as a kind of bridge, but
> a HostVirtualSwitch is more like a virtual network in libvirt terms.
>
> In terms of libvirt the interface driver is about physical NICs that
> can be used to connect a libvirt virtual network to the physical
> network. Therefore, the interface driver should just list the
> PhysicalNics, but not the HostVirtualNics. I think there is currently
> no place in the libvirt API to map the HostVirtualNics to. Also I
> don't think that it is important to make the HostVirtualNics available
> via libvirt API. I might be wrong here, but mapping them via libvirt
> interface driver is still wrong in my opinion.
>
> Regarding the network driver, I think a libvirt virtual network is
> best represented by a HostVirtualSwitch and the HostPortGroups are
> mapped to the portgroups of a libvirt virtual switch.
>
> I'm sorry that I let you wait for quite a while now until I came to
> this understanding that is contrary to your proposed interface driver.
I missed to mention that I proposed an implementation for the
interface and network drivers according to the described mapping:
https://www.redhat.com/archives/libvir-list/2012-August/msg00307.html
https://www.redhat.com/archives/libvir-list/2012-August/msg00308.html
--
Matthias Bolte
http://photron.blogspot.com
--
libvir-list mailing list
libvir-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list
Attachments:
- attachment.html (text/html — 4.7 KB)