Dan Price wrote:
On Mon 22 Jan 2007 at 11:03PM, Ben Rockwood wrote:
Still looking for a quasi-official answer on this. Again, the questions that
need specific answers to are:
A) Why, specifically, can't Non-global roots be placed on NFS?
The problem is with the attribution of the NFS traffic-- the zone root
has to be mounted over NFS *by the zone itself*-- if mounted by the
global zone, then all traffic will be interpreted as coming from the
global zone. It's important to not let NFS traffic bleed through from
one zone to another-- important enough that we have some fancy logic
to detect and reject NFS-backed pages in zone_enter(). Anyway,
it's hard to have the zone mount something if it isn't up and running.
That is to say, we have a chicken-and-egg problem.
It's not unsolvable. But it's not trivial. A second but also crucial
problem is that we will also need to be able to cope with NFS-mounted
zones at upgrade time (some of the thinking we've done around attach and
detach over the past year makes this clearer, I think).
Thank you Dan, this is the explanation I've been looking for.
B) Is anyone tasked with solving this? Is there an ARC case that I'm
The core Zones team (Steve, Jerry, and I) own the problem but we are not
currently doing work on it-- we're busy making sure the features in
Duckhorn make it into the earliest possible Solaris 10 update. We'll be
gathering in the next few weeks to sort out what we plan to tackle for
2007. We'll try to post a 2007 roadmap when we have a draft.
Let me know if you want help with anything Duckhorn related. I've
already done an internal evaluation of Jerry's putback Dec 18th and am
starting to roll it into production, even pre-SXCR B56... we just can't
wait any longer. Exciting stuff.
If you like, we can add a call record for you for the appropriate RFE.
zones-discuss mailing list