[pve-devel] multipath problems

Alexandre DERUMIER aderumier at odiso.com
Wed Aug 22 08:40:24 CEST 2012


I don't know exactly how hdparm test works. maybe this is related to thread numbers.

you can use blktace and seewatcher to see exactly how to ios are done

http://smackerelofopinion.blogspot.fr/2009/10/blktrace-baded-tools-seekwatcher.html




----- Mail original ----- 

De: "Dietmar Maurer" <dietmar at proxmox.com> 
À: "Alexandre DERUMIER" <aderumier at odiso.com> 
Cc: pve-devel at pve.proxmox.com 
Envoyé: Mercredi 22 Août 2012 08:31:03 
Objet: RE: [pve-devel] multipath problems 

> what is your hardware config ?( cpu/ram/disk). 

I just use some slow boxes to test. But I get around 150MB/s when I use a windows client, 
so that can't be the problem. 

> what's your raid level ? (raid-z, mirrors) do you use sas or sata disk? 
> 
> some quick performnce tuning,through gui: 
> :2000/settings/preferences/?this_form=sys_form 
> zfs_no_cacheflush=yes 
> Sys_zfs_vdev_max_pending = 1 for sata, = 10 for sas 
> 

I do not want to tune - IMHO there must be some fundamental bug. 

> Also if you want to have more sequential bandwith, you can try to use big 
> block size for zvol, like 128K. (I setup 4K by default in my module). 

Yes, I guess I use the wrong benchmark. Thanks for that hint. 

> seq read 
> 
> fio --filename=/dev/dm-20 --direct=1 --rw=read --bs=1m --size=20G -- 
> numjobs=200 --runtime=60 --group_reporting --name=file1 

With that I get about 153MB/s: 
# fio --filename=/dev/mapper/mpath1 --direct=1 --rw=read --bs=1m --size=20G --numjobs=200 --runtime=60 --group_reporting --name=file1 
... 
Run status group 0 (all jobs): 
READ: io=8978MB, aggrb=149921KB/s, minb=153519KB/s, maxb=153519KB/s, mint=61322msec, maxt=61322msec 

So I guess that indicates that multipathing is working correctly. 

Many thanks for your help! 

- Dietmar 



-- 

-- 



	

Alexandre D e rumier 

Ingénieur Systèmes et Réseaux 


Fixe : 03 20 68 88 85 

Fax : 03 20 68 90 88 


45 Bvd du Général Leclerc 59100 Roubaix 
12 rue Marivaux 75002 Paris 


More information about the pve-devel mailing list