On 02/04/11 12:47, GORDON MARLER, BLOOMBERG/ 731 LEXIN wrote:
Here's a few observations while using<target>:
- In the S11 Express AI Guide, Example 4-18 doesn't work as advertised.
Seems to have a typo too:
<target>
<target_device>
<swap no_swap"true"/> <!-- Should be<swap no_swap="true"/>, right? -->
</target_device>
</target>
Either way, when adding the manifest, it won't validate - complains that
<swap> needs a zvol tag, no matter what. Same issue for the<dump> tag.
This is an oversight in the dtd. To facilitate being able to specify
that one does not want a swap device with just the attribute, the zvol
subelement should not be required. I'll file a bug for this.
Note that the "no_swap" attribute should not have been documented in the
Express release. It was defined in the dtd, but not implemented yet.
For the Express release, to specify that you don't want swap (or dump)
you can add the following to your manifest:
<target>
<target_device>
<swap>
<zvol name="rpool/swap">
<size val="0"/>
</zvol>
</swap>
</target_device>
</target>
The value for the name attribute there is totally irrelevant but you
need to put something there. The size value being 0 is what tells AI
not to create swap (or dump).
- A common case we're going to be handling is where we'll be targeting disks
that either have a standard VTOC (slice 0 is very small, and other slices
exist), or were re-VTOCed sometime in the past, and the<target> is going
to have to delete all of the slices, then recreate slice 0 if you want to
dedicate the whole disk to rpool.
Here's what I finally got to work, which I thought was at least
non-intuitive. An example of this might be handy in the manual going
forward:
<target>
<target_device>
<disk>
<disk_name name="c3t1d0" name_type="ctd"/>
<slice name="0" action="delete"/>
<slice name="1" action="delete"/>
<slice name="3" action="delete"/>
<slice name="4" action="delete"/>
<slice name="5" action="delete"/>
<slice name="6" action="delete"/>
<slice name="7" action="delete"/>
<slice name="0" is_root="true" force="true"/>
</disk>
</target_device>
There's an open RFE to be able to set a whole_disk="true" attribute on
the disk element, which does what you're after here. This should be
available, along with the below, in upcoming builds.
- What's the current way to eliminate the creation of /export and /export/home?
- What's the current way to create /var as part of the BE, rather than being
shared? The idea being to set a refquota on /var so that / can grow without
bound, but /var could not.
AI in the S11 Express release doesn't support either of these
operations. This functionality is being implemented with the work being
done to support multiple pools, and should be available in upcoming builds.
-ethan
GM
_______________________________________________
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
