Slow disk performance WD5000AAKS

[jbrunink]

Hi,

Does anybody have WD5000AAKS (Western Digital) hard drive's in their Proxmox host machine? I am currently running Proxmox version 1.7, and the performance of those hard drives in my VM's and the host is very slow and changes (sometimes is the host 100+MB/s and then 12MB/s, VM's are most of the time slow)

If I run 'pveperf' I get:

CPU BOGOMIPS:      24110.82
REGEX/SECOND:      1001873
HD SIZE:           94.49 GB (/dev/mapper/pve-root)
BUFFERED READS:    77.07 MB/sec
AVERAGE SEEK TIME: 11.51 ms
FSYNCS/SECOND:     373.91
DNS EXT:           92.79 ms
DNS INT:           4.76 ms (******)

And 'pveversion'

pveversion -v
pve-manager: 1.7-11 (pve-manager/1.7/5470)
running kernel: 2.6.32-4-pve
proxmox-ve-2.6.32: 1.7-30
pve-kernel-2.6.32-4-pve: 2.6.32-30
qemu-server: 1.1-28
pve-firmware: 1.0-10
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-10
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.13.0-3
ksm-control-daemon: 1.0-4

The virtual machines have SCSI diskdrive (also tried virtio) and I benchmark the VM's and host with this command:

[root@default ~]# dd if=/dev/zero bs=1024 count=1000000 of=1gb
1000000+0 records in
1000000+0 records out
1024000000 bytes (1.0 GB) copied, 84.1533 seconds, 12.2 MB/s
[root@default ~]#

Running that command a few minutes later:

[root@default ~]# dd if=/dev/zero bs=1024 count=1000000 of=1gb
1000000+0 records in
1000000+0 records out
1024000000 bytes (1.0 GB) copied, 22.9113 seconds, 44.7 MB/s
[root@default ~]#

Dmesg of the host: http://pastebin.com/162RYAZp
Dmesg of the VM: http://pastebin.com/ZUFsFvV8

OS VM = Centos 5.5 32-bit

Can anybody give me some pointers? Or give me some advice?

Thanks.

 

[tom]

where are your virtual disk stored? I assume you just use raw files on the local storage.

power off your VM and add cache=none to you VMID.conf file and run the test again:

example (here I got ide):

ide0: vm-117-disk.raw,cache=none

 

[jbrunink]

Hi,

I have 2 HDD's in that machine. No hardware raid. I've created a LVM volume group and added that VM into that LVM volume group.

ostype: l26
memory: 512
sockets: 1
name: xerox
ide2: local:iso/CentOS-5.5-i386-netinstall.iso,media=cdrom
vlan0: virtio=66:55:47:09:77:69
bootdisk: scsi0
scsi0: data:vm-104-disk-1,cache=none
onboot: 0
cores: 1

If I run dd again I get this:

[root@default ~]# dd if=/dev/zero bs=1024 count=1000000 of=1gb
1000000+0 records in
1000000+0 records out
1024000000 bytes (1.0 GB) copied, 22.5201 seconds, 45.5 MB/s
[root@default ~]# dd if=/dev/zero bs=1024 count=1000000 of=1gb
1000000+0 records in
1000000+0 records out
1024000000 bytes (1.0 GB) copied, 24.4852 seconds, 41.8 MB/s
[root@default ~]#

Seems okay, but is this the solution? Removing cache?

Thanks your very much.

 

[tom]

yes, we think of making this as default for block devices. on the new 2.x gui we will have an option to set this.

 

[jbrunink]

The VM's are faster, but the host has now the same issue. Sometimes it's slow then it's fast... what could be the problem of that? 12MB/s seems a little bit slow.