On 02/16/11 07:08 PM, Ethan Quach wrote:
On 02/15/11 04:58, Jan Damborsky wrote:
On 02/15/11 03:09 AM, Ethan Quach wrote:
On 02/14/11 16:07, Alok Aggarwal wrote:
Hi Ethan,
On Mon, 14 Feb 2011, Ethan Quach wrote:
All,
I have uploaded the design document for AI Zones support into the
caiman-docs gate.
http://src.opensolaris.org/source/xref/caiman/caiman-docs/AI/AI_Zones_Design.odt
General review appreciated, and if so, please try to get me
comments by the end of the week; but I have also pinged
individuals to review particular parts of the document.
I took a look at 5.2.2 as requested. I just have one
small comment, the rest of the section looks okay.
Thanks for reviewing this section.
5.2.2 - The last para mentions that TI currently calls
libbe:be_init() to create an empty BE and optionally
the subordinate BE datasets and the shared datasets for it.
In reality though, TI always creates the empty BE, subordinate
datasets and shared datasets -- they aren't optional.
I meant that they are optional from the point of view of the
be_init() interface. I'll clarify this, and note TI's current usage
as well.
... but now that you've mentioned this. Somebody had recently
posted the question, "how do I get AI to not create /export and
/export/home?"
So perhaps, once we have dataset support in the manifest, TI
probably won't create all that by default, and that "default
specification" moves entirely into the default manifest.
I think there might be some coordination with zones folks needed
WRT PSARC/2010/457, as it enhanced default zone dataset layout
created by 'zoneadm install' with
<ngz_dataset>/rpool
<ngz_dataset>/rpool/export
<ngz_dataset>/rpool/export/home
mounted on /rpool, /export, /export/home within a zone respectively.
As related flag day states the last two can be destroyed if not used,
they could be perhaps treated as optional, but we might need
to check.
If this is the desired default for the 'zoneadm install' case, this
can still be dictated by some default manifest that zoneadm passes to
auto_install. So I think the underlying move of the hardcoded
datasets out of TI _could_ still be possible.
I agree. Just a nit, it's currently hard-wired in upper layer
(e.g. orchestrator), not in TI itself, though TI currently
seems to insist on being provided with at least one shared
dataset to be created.
Jan
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
