On Thu, 2014-02-20 at 15:28 -0600, Russ Anderson wrote:
> For some customers _any_ amount is significant, especially
> on large clustered systems where the amount is multiplied
> by tens or hundreds of thousands of nodes.
>
> You many not think wasting their cpu cycles is important, but they do.
Then they should be running locally built kernels in order to ensure
that there's no problematic code running at all. Distribution kernels
will always contain code that some customers won't be interested in, and
some of that code will end up executing.
If you have specific bug reports, that would be helpful. But you're not
describing actual failure conditions or showing any willingness to
figure out what the underlying problem is. The BMC is generating
messages and handing them off to the host. If that's taking more than a
tiny number of cycles, why? Once we know the answer to that, we can
actually put some effort into fixing it.
--
Matthew Garrett <matthew.garrett@xxxxxxxxxx>
��.n��������+%������w��{.n�����{�����ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
