Re: [pve-devel] [PATCH-SERIE v6 qemu-server/pve-manager] rework memory hotplug + virtiomem

["DERUMIER, Alexandre" <alexandre.derumier@groupe-cyllene.com>]

Hi Gilberto,

the idea with virtio-mem  (but also classic hotplug), is to define a
max memory.  
(currently it's max=4TB by default without any option).

Then for classic memory:

this maxmemory is shared by a number de memory slots. (64 by default),

So the increment hotplug will be = MAXmem /64 slots. (so MAX=64GB, you
can hotplug/unplug 1GB device up to 64GB,   max=256GB you can
hotplug/unplug 4GB memory up to 256GB


for virtio mem (for guest with kernel ~= >= 5.15)

it's a little bit different, because one big memory device is create,
but you can enable/disable small chunks (maxmemory / 32000).

so with max=64GB, you can hotplug/unplug 2MB increment up to 64GB, ...







Le vendredi 01 septembre 2023 à 09:36 -0300, Gilberto Ferreira a
écrit :
> Hi there.
> While I am a big zero programmer, I like to follow up these threads
> about Proxmox development.
> As always, it's a great piece of software.
> So I would like to ask if this hotplug memory will solve the problem
> we encounter when try to address huge memory in VM, like from 32 up
> to 64, without downtime.
> Much appreciated.
> 
> Thanks
> ---
> Gilberto Nunes Ferreira
> (47) 99676-7530 - Whatsapp / Telegram
> 
> 
> 
> 
> 
> Em sex., 1 de set. de 2023 às 09:24, Fiona Ebner
> <f.ebner@proxmox.com> escreveu:
> > I'll send a few patches (likely on Monday) moving a bit of code
> > around,
> > to get rid of the cyclic include
> > PVE::QemuServer <-> PVE::QemuServer::Memory
> > 
> > While it shouldn't affect your changes directly, I think it will
> > affect
> > the patch context, so you might want to wait for that before doing
> > the
> > next rebase.
> > 
> > 
> > _______________________________________________
> > pve-devel mailing list
> > pve-devel@lists.proxmox.com
> > https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
> >