[PVE-User] Slow speeds when KVM guest is on NFS

Mikhail m at plus-plus.su
Tue Nov 15 23:33:30 CET 2016


On 11/16/2016 01:22 AM, Brian :: wrote:
> Hi Mikhail
> 
> The guest that is running - what type of controller / cache?
> 
> Thanks

Brian,

The guest is Debian Jessie, running VirtIO as controller and "Default
(No cache)" cache setting. I tried both writeback / writethrough
settings as well, but it did not change things to better..

Btw, just did another "dd" test on the storage server itself (ext4
mounted from LVM lv that resides on top of MDADM RAID10):

#  dd if=/dev/zero of=test bs=64k count=550k conv=fdatasync
563200+0 records in
563200+0 records out
36909875200 bytes (37 GB) copied, 176.696 s, 209 MB/s

I guess the storage server itself is fine.

Btw, similar results when I run this test from Proxmox host that is
attached via 10gbit ethernet to storage server using NFS mount:

# dd if=/dev/zero of=test bs=64k count=550k conv=fdatasync
563200+0 records in
563200+0 records out
36909875200 bytes (37 GB) copied, 165.531 s, 223 MB/s

At this point, I don't know what else I can check on my systems to find
what's the problem with KVM images being put on NFS storage.


More information about the pve-user mailing list