[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] very slow disk access

On 2/26/2011 5:24 PM, Fajar A. Nugraha wrote:
On Sun, Feb 27, 2011 at 8:16 AM, Randy Katz<rkatz@xxxxxxxxxxxxxxxxxxxxx>  wrote:
On 2/26/2011 4:41 PM, Fajar A. Nugraha wrote:
On Sat, Feb 26, 2011 at 11:20 PM, Randy Katz
<rkatz@xxxxxxxxxxxxxxxxxxxxx>    wrote:
# lsmod | egrep "piix|hci"
ahci                   69705  7
ata1: SATA max UDMA/133 abar m2048@0xff970000 port 0xff970100 irq 253
ata1.00: ATA-8: ST31500341AS, CC1H, max UDMA/133
ata1.00: configured for UDMA/133
A quick glance seems to show no driver problems. Did you experience
any problem? Is your problem xen-specific?

I have very slow access in Xen for file based access. If I use LVM volumes
it is quite fast.
Then why didn't you say so in the first post? It shouldn't be related
to drivers. using file-based images are known to be slow or unreliable
in some circumstances.

Using the same machine with OpenVZ
has no problems and performance is great.
OpenVZ has different method to provide guest filesystem.

Will go ahead and use LVM volumes
and assume Xen likes them! :) Not sure what
else to do or look at, any comments?
If you still need file-based image, try using non-sparse image and use
tap:aio instead of file:/
Stick with LVs whenever you can though.

it seems I am using tap:aio on what appears to be quite slow:

disk = [ "tap:aio:/disk0/server3.img,xvda,w" ]

Is there a good way to convert from tap:aio to lvm? I tried stop vm, dd if=server3.img of=/dev/vg1/newlv3 but it is very very slow, much quicker to blow away the whole thing (vm) and reinstall and then add data
from backups.


Xen-users mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.