Subject: Re: [vserver] newest experimental release
From: Herbert Poetzl <herbert@13thfloor.at>
Date: Sun, 14 Sep 2008 12:04:54 +0200

On Wed, Sep 10, 2008 at 10:58:37PM +0200, diego torres wrote:
> Daniel Hokka Zakrisson escribió:
> 
> >It's always best to use whatever is newest, which happens to be 
> >2.3.0.35.5 right now.

> ok.

>>>  I can just drop a kernel patched with vs2.3.0.35 in a production
>>> machine that has 2.6.22.15-vs2.2.0.5? (after updating the userspace
>>> tools, of course)

>> The CPU scheduler is missing, and the initpid virtualization is 
>> known-broken. 

> what does mean "the cpu scheduler is missing" ? 

it means, that the fancy token bucket priority and 
hard cpu scheduler Linux-VServer normally provides
is not yet ported, and you will have to use what
mainline provides ...

> will the guests work without it?

yes, but the cpu usage limits and priorization
will not, unless you make use of cpusets, cgroup
scheduler or normal priorization mechanisms

> and what is initpid virtualization... ??? 
> self-answered in http://kerneltrap.org/node/6641

well, kind of, but it's a little more complicated
here, we are mainly talking fake init and init
less systems here

> the key is... will a vserver linux kernel with guests work 
> with this latest version?

yes, of course, and several folks already have those
versions in production

best,
Herbert

> >Depending on
> >the filesystem you use, you may also need to do a backup of your 
> >Linux-VServer flags and
> >restore those once you've booted the new kernel...
> 
> reiserfs, but backuping/restoring that shouldn't be a problem.
> 
>>> Is there any document that has some info about exp. releases? 
>>> (couldn't find any in the wiki).
>>
>> Even if there was, it would be made obsolete every other 
>> week or so... 
>> It's such a moving target, IRC is probably your best bet.
> 
> Didn't want to ask the same questions :)
>