[pve-devel] [PATCH] qemuserver : vm_status : add extended stats (disks, nics, memory).

[Alexandre DERUMIER]

> I don't have checked how you get counters for lxc. 
> We also have counters in cgroups. 

>>Yes, but different ones. 

Maybe, can we send different stats for qemu, lxc


{root}->{qemu}->{vmid}->{ballooninfo}
                      ->{blockstat}->{storageid}->{volumename} 
{root}->{lxc}->{vmid}

(Note: I would like to avoid separator characters ('_','.',..) in keys if possible,
 because graphite can be configured to used them as separator keys)



----- Mail original -----
De: "dietmar" <dietmar at proxmox.com>
À: "aderumier" <aderumier at odiso.com>
Cc: "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Mardi 16 Juin 2015 10:32:09
Objet: Re: [pve-devel] [PATCH] qemuserver : vm_status : add extended stats (disks, nics, memory).

> > + $d->{memory} = $info; 
> 
> >>Why? We already copied all data? 
> currently we send 
> 
> max_mem 
> actual 
> free_mem 
> mem = total_mem - free_mem 
> 
> ballon info return more counters 
> 
> actual 
> max_mem 
> total_mem (great to have it without doing reverse mem+free_mem) 
> free_mem 
> mem_swapped_in 
> mem_swapped_out 
> major_page_faults 
> minor_page_faults 

But you use a very generic name 'memory'. Maybe we should use 

$d->{qemu_balloon_info} = $info; 

instead? 

> >>I am quite unsure if we should directly pass qemu data. For example, I would 
> >> 
> >>like 
> >>to have qemu-server device names (instead of blockstat->{device}). 
> 
> That's a good idea. Like this if we change from virtio to scsi, we keep the 
> same stats. 
> Have the storageid too can be usefull to do aggreated stats for a specific 
> storage 
> 
> {vmid}->{disks}->{storageid}->{volumename} 

by maybe call it: 

{vmid}->{qemu_blockstat}->{storageid}->{volumename} 


> >>Also, this 
> >>only works 
> >>with qemu, so what about LXC? 
> 
> I don't have checked how you get counters for lxc. 
> We also have counters in cgroups. 

Yes, but different ones.