Thanks Sanjay. I'll update the comment.
-Drew
On 8/4/11 9:58 AM, sanjay nadkarni wrote:
My other comments have been provided/clarified in person, so the only
thing is the comment in discovery.py 653-654.
# Since libdiskmgt has no mechanism to update it's internal view of
the disk layout once its scanned /dev, fork a process and allow
libdiskmgt to instantiate the drives first. Then in the context of
the main process, when libdiskmgt gathers drive information, the drive
list will be up to date.
-Sanjay
On 07/29/11 04:56 PM, Drew Fisher wrote:
Good afternoon!
Could I please get a code review for the following CRs:
6260 <http://defect.opensolaris.org/bz/show_bug.cgi?id=6260> sparc AI need a
way to mechanism to explicitly ask for creating label on Sparc disk
7055779 <http://monaco.us.oracle.com/detail.jsf?cr=7055779> unlabeled disks
shouldn't prevent discovery from completely running
7068282 <http://monaco.us.oracle.com/detail.jsf?cr=7068282> target discovery
needs to skip removable devices such as SD cards
7071267 <http://monaco.us.oracle.com/detail.jsf?cr=7071267> discovery.py
should skip checking for iscsiadm if there's no iscsi actions
https://cr.opensolaris.org/action/browse/caiman/drewfish/format_crs/webrev/
This is a big change to the targets code so I'd like to get multiple
sets of eyes on it, especially from each of the installer folks.
Matt, Ethan, and Karen have already looked at this (and helped me
with it) but Dermot, John and Niall, could I please get you to take a
close look? The change that you need to be concerned with is in
controller.py with the new dry_run flag. The only place in the
gui-install code I could find where you used the controller was in
td_callback() and I wasn't sure how to pass a dry_run through the
callback so nothing's been changed, yet.
These changes allow us to fully utilize RFE 7037884 -
http://monaco.us.oracle.com/detail.jsf?cr=7037884 to automatically
apply a label to a disk without format asking a bunch of questions.
I have added the ability to automatically apply a VTOC label to a
totally unlabeled disk (this can only happen on SPARC due to
libdiskmgt applying a temporary VTOC label to unlabeled x86 disks).
Due to how the timing with discovery/instantiation works, the
automated labeling of an unlabeled disk must be done during discovery
so the geometry of the disk can be used by the installers. I know
this goes against the concept of discovery being practically
read-only, but I extended the dry_run flag for target discovery to
protect systems from being labeled if the user doesn't desire it.
There is an additional check on auto-labeling to prevent labeling a
disk that has a slice subelement with the 'preserve' action.
e.g.
<disk>
<slice name="0" action="preserve"/>
</disk>
will not get a label because the user explicitly told us not to.
We also now support flipping from GPT/EFI to VTOC. I do not have
support for the other directory quite yet as there hasn't been a need
for it. The only time we need an GPT/EFI label is when zfs is given
the entire disk to make a zpool and zfs handles all of that for us.
If anybody wants to play with this to see how it works, please let me
know. I can get you an ISO (x86) or can get you a SPARC machine
(thanks to Mary) with these bits.
Thanks!
-Drew
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
