[pve-devel] CPU and Memory hot add

Stefan Priebe - Profihost AG s.priebe at profihost.ag
Mon Feb 17 08:11:37 CET 2014


Am 17.02.2014 06:07, schrieb Alexandre DERUMIER:
>>> Yeah found that. But this means that the vm allocates the value of max until the balloon guest driver is loaded. So if I set it to 128gb for each guest they can't boot together as each vm tries to use 128gb.
> 
> Are you sure of that ? Because, As Far I known (for linux guest), the memory is not allocated/filled at boot, and ballonning occur when the driver is loaded very fast at boot.
> (For windows guest it's different, because it's zero filling the memory at boot)

Yes. Try the following. Pick a host with X running VMs, having may be
just 32GB free mem.

Then try to start a machine with maxmem 48GB and minimal/current balloon
value of 4GB.

It will fail...

Stefan


> 
> 
> ----- Mail original ----- 
> 
> De: "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
> À: "Alexandre DERUMIER" <aderumier at odiso.com> 
> Cc: pve-devel at pve.proxmox.com 
> Envoyé: Dimanche 16 Février 2014 10:28:27 
> Objet: Re: [pve-devel] CPU and Memory hot add 
> 
> 
> Am 16.02.2014 um 10:05 schrieb Alexandre DERUMIER <aderumier at odiso.com>: 
> 
>>>> ok so you addd maxcpus: 128 to each configto have it silently enabled? 
>>
>> Yes ;) 
>>
>> Just be carefull to set only 1 socket, as hotplug is core by core 
>>
>> (I have also send pve-manager patches in the mailing, but dietmar don't have apply them yet 
> 
> Which ones? Can't find them. 
> 
>>>> How? set in config balooning to min and max to 4096 and then changing it 
>>>> to 8192 for min max does not change anything in the guest. 
>>
>> Set max to maximum "hotpluggable" memory 
>> set min to current guest memory 
>> set shares = 0 (that disable auto ballonning from pvestatd) 
>>
>> Then play with min memory up/down 
> 
> Yeah found that. But this means that the vm allocates the value of max until the balloon guest driver is loaded. So if I set it to 128gb for each guest they can't boot together as each vm tries to use 128gb. 
> 
> Greets, 
> Stefan 
> 
>>
>>
>> ----- Mail original ----- 
>>
>> De: "Stefan Priebe" <s.priebe at profihost.ag> 
>> À: "Alexandre DERUMIER" <aderumier at odiso.com> 
>> Cc: pve-devel at pve.proxmox.com 
>> Envoyé: Samedi 15 Février 2014 17:09:39 
>> Objet: Re: [pve-devel] CPU and Memory hot add 
>>
>> Am 15.02.2014 13:59, schrieb Alexandre DERUMIER: 
>>> cpu hot add is already implemented. (even in proxmox ;) 
>>> https://git.proxmox.com/?p=qemu-server.git;a=commit;h=838776ab650f0593b48c234070c1a189fdbd72c7 
>>>
>>> cpu hot del is not yet implemented. (but they are some prelimary patches in the qemu mailing list) 
>>
>> ok so you addd maxcpus: 128 to each configto have it silently enabled? 
>>
>>>
>>> memory hotplug is not yet implemented. (should be ok for next qemu 2.0 release I think) 
>>>
>>> But maybe they are cheating, like I do : just use ballonning with a fixed size ? 
>>
>> How? set in config balooning to min and max to 4096 and then changing it 
>> to 8192 for min max does not change anything in the guest. 
>>
>> Stefan 
>>
>>
>>>
>>> ----- Mail original ----- 
>>>
>>> De: "Stefan Priebe" <s.priebe at profihost.ag> 
>>> À: pve-devel at pve.proxmox.com 
>>> Envoyé: Vendredi 14 Février 2014 21:18:47 
>>> Objet: [pve-devel] CPU and Memory hot add 
>>>
>>> Hi, 
>>>
>>> while browsing the web, i found profitbricks, which is active in germany 
>>> and the us (http://www.profitbricks.com/technical-info). 
>>>
>>> They use KVM and offer memory and cpu hot add on the fly? Anybody an 
>>> idea how this works if kvm does not already support it? 
>>>
>>> Greets, 
>>> Stefan 
>>> _______________________________________________ 
>>> pve-devel mailing list 
>>> pve-devel at pve.proxmox.com 
>>> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
>>>



More information about the pve-devel mailing list