On 09/05/2012 04:55 PM, Yu Mingfei wrote:
On 28/08/12 18:28, Yu Mingfei wrote:
On 28/08/12 15:05, Alex Jia wrote:
On 08/27/2012 09:55 AM, Yu Mingfei wrote:
On 24/08/12 17:54, Alex Jia wrote:
Hi Mingfei,
This test cases just check virsh command output and haven't touch any actual cgroup relevant parameters check, you know libvirt uses cgroup 'cpu' controllers
to make use of with QEMU guests, so you should check them in cgroup.

Thanks, Alex.
It is a great idea, and I will have a investigation on it.

But because this is a rudiment of schedinfo, I may still just check output of it. If it is necessary, I will replace it with actual cgroup relevant parameters check
after my investigation.^^



Note that is we shouldn't check virsh command by itself unless we can't find other methods. So the previous case is incomplete if it hasn't check cgroup relevant parameters.

Regards,
Alex


Thanks Alex:

In fact, I'm not so clear about cgroup, but according your advice, I'm investigating it.
And I will give a reply as soon as possible^^.


Hi Alex, Chris:

With roughly understand of cgroup, I can get cpu_shares in cpu controller by two ways:

1.cgroup tools:
# cgget -n -r cpu.shares libvirt/qemu//vm1/ on RHEL6
and
# cgget -n -r cpu.shares system/libvirtd.service/libvirt/qemu//vm1/ on fedora17

Right way.


2.access the file cpu.shares in cpu controller directly.

But I find the cpu.shares of libvirt in cgroup is different on fedora and RHEL6, just as up command: (xxx/cpu/libvirt/qemu/ on RHEL6 while xxx/cpu/system/libvirtd.service/libvirt/qemu/ on fedora).

You may parse 'lscgroup'(from libcgroup-tools rpm) output then get a actual cgroup path, but I'm not sure 'lscgroup' is always available for all of platforms.

Alex


I prefer cgroup tools to get cpu.shares because it will be easier.:-)
Please tell me If I have some omission.

thanks.
--
Best Regards
Yu Mingfei

_______________________________________________
Autotest-kernel mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/autotest-kernel

Reply via email to