>>> Then it is the same situation if you define a cgroup in cpuset in 
>>> cgconfig.conf. I plan to do examples and document the configuration nicely.
>>> From my point this is the best solution. Is it ok for you?
>>
>> I would prefer a sane default, such as copy_from_parent, which means,
>> we will almost always have a sane default, which works.
>
> There is no sane default without extensive logic in libcgroup, copy from
> parent won't work on cpuset with cpu_exclusive.
>

Correct, but I will argue that for most cases, cpu_exclusive is not
used, and people expect it to work like they expect other cgroups to
work (that is inherit sane values from parent).

Let me put it this way, if you are using cpu_exclusive, and using
libcgroup at the same time, you oughta know what cpusets are, and how
to configure libcgroup. And you are expected to either turn of
automatic classification, or configure it with a sane default for your
use case.

However, not using cpu_exclusive, might mean that you do not know
about cpusets, and do not care about them, (And if you do, then the
same caveat that applied in the previous case applies to you). In this
case, we _have_ to provide sane defaults which is, use all the memory,
use all the CPUs.

> Current cgconfigparser does simple mkdir for groups, which are not
> configured. It is up to the administrator to pre-configure parent groups
> if simple mkdir is not enough.
>
> I think that cgrulesengd should do exactly the same.

This is where I do not agree. I believe we made a mistake originally
by not setting sane defaults. It has bitten us more than once. While I
agree, it is hard to come up with that, we do need both policy as well
as mechanism to be setup.

Dhaval

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Libcg-devel mailing list
Libcg-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libcg-devel

Reply via email to