On 06/29/10 04:10 PM, Sue Sohn wrote:
On 06/24/10 05:15, Matt Keenan wrote:
Hi,
Thanks for the feedback/review, I've incorporated changes back into
the design document and a
2nd draft of Automated Installer Multi Disk/Pool support design is
now available
for review :
Version 0.2 :
http://hub.opensolaris.org/bin/download/Project+caiman/auto_install/ai%2Dmulti%2Ddesign%2D0.2.pdf
Cheers
Matt
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
Hi Matt,
2.1 lib/libtarget_pymod section is listed twice
Fixed
2.2.1
o <iscsi_target_paramaters> -> <iscsi_target_parameters>
Fixed
o <target_device_type> / <target_device_vendor> / <target_device_sizeice>
target_device_sizeice -> target_device_size
Fixed
o "Can only specify one <ai_target_device> element."
If more than one ai_target_device is specified in the manifest, is the
manifest considered invalid? This should be indicated in the doc.
This section of the Document was just an analysis of the existing
ai_manifest.rng schema, and as this schema is being obsoleted by the new
AI schema design I don't think there's a need to specify any further
information.
o "If not specified in AI manifest, default is to select current boot
disk, otherwise the first disk with sufficient size for install is
chosen."
Back during some earlier AI client design work, I believe that Frank
said that if no disk selection specification had been made in the
manifest, then the manifest should be considered invalid. This may
have changed, but would be good to check with him if you haven't already.
If it is decided to leave this as is, how is the "first disk"
determined? Will this provide repeatable behavior (i.e., will the same
disk always be chosen in this situation)?
I did a little bit of more digging into the code, and what's stated
above is not entirely correct. I've changed the statement to :
*
Target device name can be either a disk name c#t#d# or boot_disk.
If not specified the default is to select the first disk that
matches the minimum size requirements.
To select the current boot disk by default the target_name needs to be
"boot_disk". If no target is specified at all, then the first disk that
matches criteria in auto_td.c:select_default_disk() e.g. minimum size is
chosen. As to repeatable behaviour, the disks are processed in the order
returned by libtd. So if libtd always returns the list of discovered
disks in the same order then yes it is repeatable.
o if multiple disks meet the device specification (e.g., multiple
disks meet target_device_size or target_device_vendor), how is it
determined which disk to use?
The first disk that is found in libtd supplied sequence is chosen.
2.2.4 has been down -> has been done
Fixed
3.2.1.1
2nd bullet: Optional Attribute "root"
I think this should be "is_root" to match what is in Appendix 1.
5th bullet: property-value -> property=value
6th bullet: target/slice/patition -> target/slice/partition
3.2.1.3 The second ai_device_vtoc_slices should be
ai_device_vtoc_slice_contents
I've removed all the proposed schema design as all schema related design
is now contained in the AI/DC schema redesign proposal, to which a
reference will be available in this design doc.
3.2 You mentioned libtarget_pymod and libti_pymod in section 2.1, but
though you expand here on the other consumers mentioned there, such as
libti and liborchestrator, don't see anything else about
libtarget_pymod or libti_pymod. Should they be included in this
section too?
Good point, I've added them.
thanks
Matt
Thanks,
Sue
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
