[libvirt] [PATCH] blkio: change the minimum weight from 100 to 10
Gao feng
gaofeng at cn.fujitsu.com
Wed Aug 21 08:24:48 UTC 2013
On 08/20/2013 01:23 AM, Eric Blake wrote:
> On 08/18/2013 11:59 PM, Gao feng wrote:
>> kernel had changed the minimum weight of device blkio from
>> 100 to 10 in commit df457f845e5449be2e7d96668791f789b3770ac7.
>>
>> commit df457f845e5449be2e7d96668791f789b3770ac7
>> Author: Justin TerAvest <teravest at google.com>
>> Date: Tue Mar 8 19:45:00 2011 +0100
>>
>> blk-cgroup: Lower minimum weight from 100 to 10.
>>
>> We've found that we still get good, useful isolation at weights this
>> low. I'd like to adjust the minimum so that any other changes can take
>> these values into account.
>>
>> Signed-off-by: Justin TerAvest <teravest at google.com>
>> Acked-by: Vivek Goyal <vgoyal at redhat.com>
>> Signed-off-by: Jens Axboe <jaxboe at fusionio.com>
>>
>> libvirt should comport with kernel.
>>
>> Signed-off-by: Gao feng <gaofeng at cn.fujitsu.com>
>> ---
>> docs/formatdomain.html.in | 4 ++--
>> src/util/vircgroup.c | 10 +++++-----
>> 2 files changed, 7 insertions(+), 7 deletions(-)
>
> What happens when running a newer libvirt with an older kernel? Or in
> other words, what error message do you get if you pass a limit lower
> than the current kernel can support? I want to make sure the message
> looks sane to an end-user before accepting this patch.
>
>>
>> - if (weight && (weight > 1000 || weight < 100)) {
>> + if (weight && (weight > 1000 || weight < 10)) {
>> virReportError(VIR_ERR_INVALID_ARG,
>> - _("weight '%u' must be in range (100, 1000)"),
>> + _("weight '%u' must be in range (10, 1000)"),
>> weight);
>> return -1;
>
> In other words, I suspect this code needs to be beefed up to actually
> probe whether the kernel accepted the change, rather than blindly doing
> the filter ourselves and hoping that it was correct.
>
Hmm, I haven't considered this problem. we should report different error messages
and check different min values base on different kernel.
Thanks!
More information about the libvir-list
mailing list