Peter:
Thanks for the comments.
Peter Tribble wrote:
> On 3/22/07, Sarah Jelinek <Sarah.Jelinek at sun.com> wrote:
>> Hi All,
>>
>> I posted new design documents for the Dwarf Orchestrator and Dwarf
>> Target discovery pieces on the Dwarf project page. These are
>> significantly different than the first drafts of the designs.
>>
>> If possible, please review and comment by COB 4/6/07:
>>
>> http://opensolaris.org/os/project/caiman/Dwarf/#Dwarf_Caiman_Design_Documents
>>
>>
>
> Looking at the orchestrator document.
>
> Section 4.1 - Disk information
>
> Why is the type of disk and manufacturer important?
We want to display that information to the user. The current installer
just says c0d0, which is not that useful.
>
> Section 4.1 - VTOC information.
>
> Are you making use of the Tag? This may have had some significance at
> some time in the past, but I'm not sure anything or anyone uses it any
> more.
>
> The same is true, to a lesser extent, to the Flag.
We just want to capture all the available information. The current
installer uses tag to determine whether the slice (if it is root) needs
to be examined for existing Solaris instances for upgrade.
>
> Looking at the data structures:,
>
> diskInfo:
> diskSize is a long - what are the units? (Surely not GB as the comment
> indicates?)
Should be MB.
>
> SlieInfo_t:
> sliceSize - again, what are the units (Comment now says MB.)
MB
>
> ufsInstance:
> svmConfigured - the comment says ufs root, is this just ufs root or
> that it's
> a member of any metadevice.
This is a boolean that says whether this particular Solaris instance is
configured as a SVM root mirror.
>
> I don't see anywhere where metadevice databases are recorded.
> Do they need to be?
For upgrade, we only need the name of the disk slice where Solaris
instance is located. We are looking for SVM mirror root because we don't
want to show all submirrors to the users since upgrading the primary
upgrades all other submirrors of the mirror device. The backend code
needs only the name of the primary slice and it will determine the
metadbs and other slices.
- Sundar
>
> In 51., for sysidtool, shouldn't timeserver just be set to localhost?
>
> Oh, and on page 8 of the TD doc, it should be scsi not scsci.
>
