On Tue, 26 Jan 2010, Ethan Quach wrote:

>> auto-installer: line 124: Where is SMF_FMRI defined?
>> (I don't think it is). Also, do you want it to be
>> AI_SMF_FMRI instead?
>
> SMF_FMRI is set as the running method's service fmri.

Okay, I learned something here :)

>> manifest-locator: line 243: Remove the use of $HEAD, yields
>> the same results.
>
> for cases where the same bootarg is given more than once,
> and we're not expecting it to be, the processing becomes faulty.
> For example, if there existed boot properties "...,foo=bar,...,...,foo=bar"
> the value for 'foo' ends up being "bar bar".  We're not expecting
> multiple instances of any of our boot properties so I figured taking
> the first instance is a better compromise, over errorring out.

Okay.

>> manifest-locator: line 302: I trust you've reworked this piece
>> so the correct check happens for x86?
>
> I've turned the check for the ramdisk boot block device to be Sparc
> specific, and added an x86 specific check to look for the 'install_media'
> boot property.  If that exists, its assumed to be a net boot.  (I'm trying to
> see if there's a better way to determine that we've booted from the
> network on x86, but there doesn't really seem to be.)

Sounds good.

>> dc_defs.py: Why bother setting the do_safe_default keyword in
>> DC and propagate it via the .image_info file? Why not just
>> always create a safe default GRUB menu entry especially since
>> we don't provide a user visible way to tweak DC to not be
>> the case?
>
> So that there's still that separation from what decides when to do it.
> For example, if we want an 2010.03 Ai server to properly support
> 2009.06 Ai images we need to do it this way -- embed the decision in
> the image.  The same reasoning going forward if future images decide
> to change things in some way I guess.

I think that may work for the immediate case but longer
term this sounds like an ad-hoc solution from an extensibility
point of view. The issue of maintaining multiple image versions
on the AI server is a generic one, I'm not sure if solving it
by embedding tags into the image_info file is the correct approach.

> On the DC end of things, I wasn't sure I'd wanted to make it a tunable
> or if it should be a tunable just to make the impl cleaner, and not reliant
> on the build# of the build system.  Thoughts?

I'm ambivalent to exposing this via the DC interfaces. I don't
see an advantage of doing so offhand.

>> installadm.1m.txt: Do we also want to document here how to change
>> the default GRUB menu entry to do a hands free install?
>
> I don't know if that's a good idea.  We're trying to prevent something
> bad here, so I'd rather not note that in the manpage.  AIs been behaving
> this way up til now, but that doesn't mean it should have been. IMO if
> someone really wants to do this, they'll figure it out.

Sounds reasonable.

Alok

Reply via email to