Joe Pawlicki wrote:
> Again, maybe something so basic that everyone knows it, but I don't see any
> reason why you can't run zones on a system under the FSS, for some pretty
> functional load balancing, without having to create and associate pools.
> Just configure your zone.cpu-shares in each zone, enable the FSS in the GZ,
> and that's it.  I put a load on each zone and verified with "prstat -Z",
> and it works as expected.
> Yet every example I've come across so far, including 
>  makes it look like it's necessary to set up and associate pools with each
> zone.
> Granted, this is a fairly coarse control, and just ensures that each zone
> gets its fair share, not apps within each zone.  But for something like my
> immediate needs, like creating dev sandboxes where each engineer gets his
> own zone, or putting a couple different app/web servers on a box, isn't the
> approach above sufficient?  

For your needs FSS should be sufficient.  In fact, FSS is sufficient for most 
purposes, and is the general recommendation.  It might even become the default 
some day.

Few applications need their CPU usage to be capped.  Capping usually leads to 
wasted CPU cycles.  But there are situations where capping is the right 

> Or am I once again too green at this to see the
> obvious reason why all the examples include pool setup?

Perhaps most examples use pools because, until Solaris 10 8/07, configuring 
them is more complicated than FSS, and therefore most people writing examples 
see more value in describing the use of pools.  How many examples of "ls" 
usage do you see featured in blogs? :-)

Jeff VICTOR              Sun Microsystems            jeff.victor @
OS Ambassador            Sr. Technical Specialist
Solaris 10 Zones FAQ:
zones-discuss mailing list

Reply via email to