[PVE-User] vzdump.conf settings not taken into account

[Luca Fornasari]

-------- Original Message --------
Subject: Re: [PVE-User] vzdump.conf settings not taken into account
From: Martin Maurer <martin at proxmox.com>
To: pve-user at pve.proxmox.com <pve-user at pve.proxmox.com>
CC:
Date: 02/21/2012 11:14 AM
>> -----Original Message-----
>> From: pve-user-bounces at pve.proxmox.com [mailto:pve-user-
>> bounces at pve.proxmox.com] On Behalf Of Luca Fornasari
>> Sent: Dienstag, 21. Februar 2012 10:52
>> To: pve-user at pve.proxmox.com
>> Subject: Re: [PVE-User] vzdump.conf settings not taken into account
>>
>> Martin,
>>
>> I mean: to have good vzdump performances (speed) which disk is the most
>> important? The originating disk where VMs reside or the destination disk where
>> vzdump creates the .tar.gz?
> [Martin Maurer]
>
> Depends on many factors. If you use gzip compression (tar.gz) also your cpu is a limiting factor (only one cpu is used here). Check your processes during the backup process. With 2.0 rc1, we introduced lzo compression, much faster and the default now.
> If you store your backups on a NAS, obviously also the network has impact.
>
> For most it makes no sense to put all resources of the host and network to speed up backups. Much more important is a low load during backups in order that all VM´s are working well during backups.
>
> Martin

As an example: I was doing backups on an external box using NFS and no 
gzip compression; 1Gbit/sec ethernet link. The external box was running 
linux and software based RAID5 on cheap SATA disks. When the NFS daemon 
on the external box was in D state because the SATA disks were slow 
(atsar was showing 100% utilization) a *random* KVM virtual machine (on 
top of LVM) on the node was going on D state as well.

So I had to decrease bandwidth in vzdump.conf

What was not making sense to me is: why a slowness on the destination 
disks was decreasing performances on the source disk?

Any ideas ?

Best Regards,
Luca