[pve-devel] Hyperconverged Cloud / Qemu + Ceph on same node

[Stefan Priebe - Profihost AG]

Am 23.07.2018 um 21:04 schrieb Alexandre DERUMIER:
> Personally, I think that a vm could take all cpu usage, or memory, and impact ceph cluster for other vms.
> 
> we should give ceph some kind of (configurable) priority.

Yes the problem is or might be. Most users will overcommit the cpu cores
with VMs. So theoretically the kvm processes might eat ALL CPU cores.
But while running ceph on the same node this should NEVER happen. So my
idea is to give cores exclusively to ceph.

Another problem is NUMA currently we use single CPUs to prevent all the
NUMA mess for ceph. So if i want to run ceph on Dual Socket Machines i
also want to pin ceph exclusively to cores on the same CPU.

Greets,
Stefan

> ----- Mail original -----
> De: "dietmar" <dietmar at proxmox.com>
> À: "pve-devel" <pve-devel at pve.proxmox.com>, "aderumier" <aderumier at odiso.com>
> Envoyé: Lundi 23 Juillet 2018 20:53:19
> Objet: Re: [pve-devel] Hyperconverged Cloud / Qemu + Ceph on same node
> 
> I am not sure CPU pinning helps. What problem do you want to solve 
> exactly? 
> 
>> maybe could we use cgroups ? (in ceph systemd units) 
>>
>> we already use them fo vm && ct (shares cpu option for priority) 
>>
>>
>> ----- Mail original ----- 
>> De: "Stefan Priebe, Profihost AG" <s.priebe at profihost.ag> 
>> À: "pve-devel" <pve-devel at pve.proxmox.com> 
>> Envoyé: Lundi 23 Juillet 2018 13:49:17 
>> Objet: [pve-devel] Hyperconverged Cloud / Qemu + Ceph on same node 
>>
>> Hello, 
>>
>> after listening / reading: 
>> https://www.openstack.org/videos/vancouver-2018/high-performance-ceph-for-hyper-converged-telco-nfv-infrastructure 
>>
>> and 
>> https://www.youtube.com/watch?v=0_V-L7_CDTs&feature=youtu.be 
>> and 
>> https://arxiv.org/pdf/1802.08102.pdf 
>>
>> I was thinking about creating a Proxmox based Cloud with Ceph on the 
>> same nodes as proxmox. 
>>
>> What i'm missing is HOW do i get automatic CPU pinning for Qemu and 
>> Ceph? How can they live in parallel without manually adjusting cpu 
>> pinning lists? Has anybody already tried it? 
>>
>> Greets, 
>> Stefan 
>> _______________________________________________ 
>> pve-devel mailing list 
>> pve-devel at pve.proxmox.com 
>> https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
>>
>> _______________________________________________ 
>> pve-devel mailing list 
>> pve-devel at pve.proxmox.com 
>> https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
> 
> _______________________________________________
> pve-devel mailing list
> pve-devel at pve.proxmox.com
> https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
>