Darren and Dermot,
I'm perfectly fine with removing the extended_partition entry based on
what you guys said. I didn't realize (and will add a much needed
comment to the DTD file explaining) the convention of 1-4 being primary
and 5-36 being logical.
Are both of you ok with moving the in_zpool/in_vdev attribute up to the
<disk> level?
-Drew
On 1/25/11 6:51 AM, Darren Kenny wrote:
I tend to agree with Dermot here, specifically for the case where we want do to
something like:
<partition action="use_existing_solaris2"/>
If we differentiate between logical and primary in the schema, then the above
wouldn't work as expected where the Solaris2 partition is a logical partition.
Also, the 5-36 notion is also how fdisk itself interprets things with its -W and
-F arguments, so we would also be deviating from that "standard" in Solaris.
(Okay, there aren't numbers, but it's done by position in the output/input of
fdisk here).
Thanks,
Darren.
On 25/01/2011 13:38, Dermot McCluskey wrote:
Drew,
I'm a bit concerned that such a big change is coming in at the
end of our discussion.
As I understand it, the way this is meant to work is that the partition
name is used to identify whether a partition is logical or primary,
ie 1-4 = primary; 5-36 = logical. And there can only be one extended
partition per disk, so it is always possible to determine which extended
partition the logical partitions belong to.
Admittedly, it might appear a bit neater if the logical partitions were
sub-elements of the extended partition in the schema, but even so, it
should definitely be possible to specify legitimate layouts with the
existing DTD.
Are there any specific use-cases that cannot be accommodated by the
current schema?
- Dermot
btw, why would you need a part_type attrib on extended_partition?
What values could it possibly have other than "15"?
On 01/25/11 03:03, Drew Fisher wrote:
Oh hey.
Good catch.
I'll dump the in_zpool/in_vdev attributes and the logic in logical.py
tomorrow morning.
Thanks Niall!
-Drew
On 1/24/11 6:47 PM, Niall Power wrote:
Hi Drew,
This one caught my attention:
<!ELEMENT extended_partition (partition*, size?)>
<!ATTLIST extended_partition name CDATA #IMPLIED>
<!ATTLIST extended_partition action
(create|delete|preserve|use_existing_solaris2) "create">
<!ATTLIST extended_partition part_type CDATA "15">
<!ATTLIST extended_partition in_zpool CDATA #IMPLIED>
<!ATTLIST extended_partition in_vdev CDATA #IMPLIED>
Reading this, it seems to suggest that an extended partition can be a
member of a zpool or a vdev composition. I can't see how this is
possible. An extended partition should only be directly
usable as a container for logical partitions.
I would expect a logical partition that is a sub-element of the
extended_partition (represented by "partition*") to be the only thing
within that could form part of a Zpool or Vdev composition.
Thanks,
Niall
_______________________________________________
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
