[Engine-devel] host cpu feature
Yaniv Kaul
ykaul at redhat.com
Wed Dec 5 14:05:00 UTC 2012
On 12/05/2012 03:55 PM, Laszlo Hornyak wrote:
>
> ----- Original Message -----
>> From: "Yaniv Kaul" <ykaul at redhat.com>
>> To: "Laszlo Hornyak" <lhornyak at redhat.com>
>> Cc: "Dan Kenigsberg" <danken at redhat.com>, "engine-devel" <engine-devel at ovirt.org>
>> Sent: Wednesday, December 5, 2012 2:45:46 PM
>> Subject: Re: [Engine-devel] host cpu feature
>>
>> On 12/05/2012 03:39 PM, Laszlo Hornyak wrote:
>>> ----- Original Message -----
>>>> From: "Dan Kenigsberg" <danken at redhat.com>
>>>> To: "Laszlo Hornyak" <lhornyak at redhat.com>
>>>> Cc: "Yaniv Kaul" <ykaul at redhat.com>, "engine-devel"
>>>> <engine-devel at ovirt.org>
>>>> Sent: Wednesday, December 5, 2012 1:55:19 PM
>>>> Subject: Re: [Engine-devel] host cpu feature
>>>>
>>>> On Wed, Dec 05, 2012 at 06:46:09AM -0500, Laszlo Hornyak wrote:
>>>>> ----- Original Message -----
>>>>>> From: "Yaniv Kaul" <ykaul at redhat.com>
>>>>>> To: "Laszlo Hornyak" <lhornyak at redhat.com>
>>>>>> Cc: "engine-devel" <engine-devel at ovirt.org>
>>>>>> Sent: Wednesday, December 5, 2012 12:23:47 PM
>>>>>> Subject: Re: [Engine-devel] host cpu feature
>>>>>>
>>>>>> On 12/05/2012 12:32 PM, Laszlo Hornyak wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> CPU-Host support allows the virtual machines to see and utilize
>>>>>>> the
>>>>>>> host's CPU flags, this enables better performance in VM's, at
>>>>>>> the
>>>>>>> price of worse portablity.
>>>>>>>
>>>>>>> http://www.ovirt.org/Features/Cpu-host_Support
>>>>>>>
>>>>>>> Your feedback is welcome!
>>>>>>>
>>>>>>> Thank you,
>>>>>>> Laszlo
>>>>>>> _______________________________________________
>>>>>>> Engine-devel mailing list
>>>>>>> Engine-devel at ovirt.org
>>>>>>> http://lists.ovirt.org/mailman/listinfo/engine-devel
>>>>>> - I assume that when you allow migration, you'd use host-model?
>>>>>> This
>>>>>> is
>>>>>> not clear from the design. It seems like we VDSM developers can
>>>>>> choose
>>>>>> to use either this or passthrough, while in practice we should
>>>>>> support both.
>>>> I join Kaul's question: it is an ovirt-level question whether
>>>> hostPassthrough or hostModel or both should be supported. It
>>>> should
>>>> not
>>>> be a unilateral vdsm decision.
>>> Ah, possibly misunderstanding, I did not mean that VDSM should
>>> decide whether to use host-passthrough or host-model. The engine
>>> should decide.
>>> I meant _you_ should decide which version of vdsm api modification
>>> do you want :)
>>>
>>>>> If AllowMigrateCPUHost is set to true (in case you have the same
>>>>> cpu model everywhere in your DC) migration of such hsots will be
>>>>> enabled. Otherwise it will not be enabled.
>>>> What is the breadth of AllowMigrateCPUHost? Engine wide? Per DC?
>>>> Per
>>>> cluster?
>>> I thought of eninge-wide. The of course you can have different
>>> models in two different DC, but they should be unique in one.
>>> We can add this to DC or cluster level, imho it would be just
>>> another checkbox on the UI that most users would not use.
>>>
>>>> I favor the latter; a user may have a cluster of exact-same hosts,
>>>> where
>>>> hostcpu migration is allowed, and other cluster where it is
>>>> forbiden.
>>>>
>>>> The nice thing about hostModel (unlike hostPassthrough) is that
>>>> once
>>>> we
>>>> created the VM we can migrate it to stronger hosts, and back to
>>>> the
>>>> original host. I suppose that it complicates the scheduler.
>>> Yes with host-model you get the features that libvirt handles. In
>>> such cases the engine could decide, if you want this
>>> functionality. Well the scheduler architecture is just being
>>> reinvented.
>>>
>>> For the host-passthrough, I think the AllowMigrateCPUHost
>>> configuration option would be a simple decision for the
>>> administrator: set it to true if all hosts are uniform. If it is
>>> not set to true, then we will not allow migration of such VMs.
>> That's not what I understood from libvirt's documentation. I
> You may be right, could you send an URL to that point of the documentation or copy-paste?
The link I followed from your feature page:
http://libvirt.org/formatdomain.html#elementsCPU :
host-model
The host-model mode is essentially a shortcut to copying host CPU
definition from capabilities XML into domain XML. Since the CPU
definition is copied just before starting a domain, exactly the same XML
can be used on different hosts while still providing the best guest CPU
each host supports. Neither match attribute nor any feature elements can
be used in this mode. Specifying CPU model is not supported either, but
model's fallback attribute may still be used. Libvirt does not model
every aspect of each CPU so the guest CPU will not match the host CPU
exactly. On the other hand, the ABI provided to the guest is
reproducible. During migration, complete CPU model definition is
transferred to the destination host so the migrated guest will see
exactly the same CPU model even if the destination host contains more
capable CPUs for the running instance of the guest; but shutting down
and restarting the guest may present different hardware to the guest
according to the capabilities of the new host.
host-passthrough
With this mode, the CPU visible to the guest should be exactly the same
as the host CPU even in the aspects that libvirt does not understand.
Though the downside of this mode is that the guest environment cannot be
reproduced on different hardware. Thus, if you hit any bugs, you are on
your own. Neither model nor feature elements are allowed in this mode.
Y.
>
>> understood
>> that if you want host+migration, you need to use host-model.
>> Otherwise -
>> host-passthrough.
>> Y.
>>
>>>>>> - I'm still convinced and commented on both relevat oVirt and
>>>>>> libvirt
>>>>>> BZs that we need to add x2apic support to the CPU, regardless of
>>>>>> what
>>>>>> the host CPU exposes.
>>>>>> AFAIK, the KVM developers agree with me.
>>>>> Not quite sure how is this related... could you send some URL's
>>>>> for
>>>>> the bugreports?
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/engine-devel/attachments/20121205/8aee88d9/attachment.html>
More information about the Engine-devel
mailing list