Dan Price wrote:
> On Wed 14 May 2008 at 03:15PM, Mike Gerdts wrote:
>> On Wed, May 14, 2008 at 3:06 PM, Jordan Brown <[EMAIL PROTECTED]> wrote:
>>> I remember hearing that IPS was designed to work well in a Zones
>>> environment.  Can anybody point me at any details?
>> http://blogs.sun.com/dp/entry/a_field_guide_to_zones
> 
> To be clear: the design takes zones into account-- but we're not
> yet to the stage of fully exploiting that design for zones.

Hmm.  Interesting reading, but it seems like you haven't really gotten 
into the interesting parts.

Sparse-root zones are one of the key pain points.  Eliminating them 
makes life easier... but then again, if we could eliminate them, life 
would be easier for SVR4 packaging too.  As a person developing an 
application for deployment onto a Zones system, sparse-root zones are 
probably my biggest Zones headache by far.

Zones aren't completely independent of the global zone.  Better not 
install a libc that's newer than the kernel that's running.

One of the key advantages of zones over multiple systems or over VMs is 
that in many cases you can ignore them and just manage the global zone, 
and the local zones will just take care of themselves.  The more that 
they differ from one another, the less that works and the more complex 
the management processes are.  Eventually you end up with something that 
looks like a multi-system management infrastructure.  Those work, but 
they tend to be pretty big and complicated, considerably more complex 
than having "pkgadd" update all of the zones at once.

Being willing to change the basic assumptions is a good thing, and 
usually one of the only ways to get serious improvements.  Just make it 
clear how much baby is in the bathwater.
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to