Xen users - Sep 2006 - RE: How to configure VBD hard disk drives in Linux HVMdomain?

>-----Original Message-----
>From: xen-users-bounces@lists.xensource.com
>[mailto:xen-users-bounces@lists.xensource.com] On Behalf Of Liang Yang
>Sent: 2006年9月19日 8:36
>To: xen-users@lists.xensource.com
>Subject: [Xen-users] How to configure VBD hard disk drives in Linux
>HVMdomain?
>how to do it in the config file?
Your config is partly right. But it also needs a Qemu emulated IDE disk as the
OS to boot up firstly, for example, ''hda''. It is because HVM
VBD disk can be used, only after HVM boot up and load VBD driver. So the config
could look like:
disk = [ ''''phy:/dev/sda,hda,w'' and
‘phy:/dev/sdb,xvda,w’]
>I tried the following formats:
>disk = [''phy:/dev/sda,xvda,w'' and ‘phy:/dev/sdb,xvdb,w’]
>
>but these drives are still not recognized. Do I need to build
>para-virtualized driver before I can use vbd in Linux HVM?
Yes. Correct.

Compared with configuration, the more operations are need before and after
creating HVM. Usually it needs to rebuild HVM kernel and HVM VBD driver firstly.
I did following ways (It is okay in xen unstable changeset 11440):
1. Download and build 2.6.16 Linux kernel as HVM new kernel. (If you HVM guest
is UP, you need do extra config. The build process can be operated in Xen0)
2. Find you Xen0 source. Go to folder unmodified_drivers/linux-2.6 . Follow
README to build out 4 HVM VBD drivers.
3. Install new kernel (step 1) in HVM and copy 4 VBD drivers to HVM.
4. Create HVM with the former config and the new kernel. 
5. After HVM is up, follow ''step 2'' README to load VBD drivers
one by one.
6. If everything is okay. ''fdisk -l'' could find the new hard
disk ''xvda''. You can mount it to any folder to use it. ;)

If any problems meet, please let me know. ;)

Best Regards,
Yongkang (Kangkang) 永康

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Hi,

I did some SAS physical drive performance testing using IOMeter. I used the 
latest xen-unstable version with Linux kernel 2.6.16.13. The Linux I used is 
Red Hat Enterprise AS 4.4. I compared the performance data between Xen Linux 
Domain0 and Xen Para-Virtualized Linux Guest DomainU. I used the same Linux 
kernel and ramdisk module for both dom0 and domU. To my surprise, I found 
Linux domainU has better physical drive performance than Xen Linux domain0 
(for both small packet size and large packet size).

Can anyone explain this?

Thanks,

Liang


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Hi,

I originally built Xen on a SMP platform. Later I need to run some testing 
on Uni-Processor platform. So I append nosmp to Xen boot options. Now Xen 
can boot into a Uni-Processor mode. After doing some I/O performance 
benchmarking, I really doubt if there is performance impact by adding nosmp 
and force a SMP version of Xen to a UP version. Is it better to use Xen 
which is built on a native Uni-Processor platform instead of using nosmp to 
a SMP Xen version?

Thanks,

Liang

 


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Hi,

I''m running some testing on Xen 3.0.3. The platform I used has four
cores
(two dual-core Intel Xeon CPUs). I try to assign just one VCPU to Xen 
domain-0. After doing the command "xm vcpu-set Domain-0 1", now the
"xm
list" command shows only 1 VPCU is indeed assigned to domain-0. The command
"cat /proc/cpuinfo" also shows there is just one CPU in the system.
However,
when I cat the information from /proc/stat, I still find four CPUs are 
listed there.

Could someone explain the reason for this inconsistent information?

Thanks,

Liang


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users