On 19/01/2011 12:05, Matt Keenan wrote:
> On 01/19/11 11:54 AM, Darren Kenny wrote:
>> On 18/01/2011 21:21, Dave Miner wrote:
>>> On 01/18/11 11:22 AM, Drew Fisher wrote:
>>>> Dermot,
>>>>
>>>>
>>>> On 1/18/11 7:53 AM, Dermot McCluskey wrote:
>>> ,,,
>>>>> - Q: where do<swap> and<dump> go? under<target> or<logical>?
>>>>>
>>>>
>>>> Honestly, I don't know yet. For Zvol based swap/dump, perhaps that
>>>> could be an attribute to a<zvol> object which is a child under a
>>>> <zpool> object? is_swap (true|false) "false" or something like that?
>>>>
>>>> For UFS based ... I have no idea. Are we even supporting UFS swap/dump?
>>>>
>>>
>>> There are some corner-case advantages to swap and dump slices (don't say
>>> UFS, because it doesn't involve UFS :-) over zvol's at present that
>>> might appeal to specific situations. If we can easily support them in
>>> the syntax, that would be good, even if we don't implement support in
>>> the code immediately (or ever).
>>>
>>
>> I believe the last time we met (Friday) to talk about targets, we talked
>> about
>> adding a 'swap=(true|false}' attribute to the<slice> tag, defaulting to
>> 'false'
>> - and I still feel this is the best approach to allow creation of a swap
>> slice.
>>
>> This would handle the swap on a slice or not.
>>
>
> How would one enforce swap/dump to not be created ?
In the case of swap slice, just don't specify any slices with "swap=true"!
But in the case of a zvol based swap, then that's different, and in that case
its a logical element like:
<swap no_swap="true"/>
(The currently target.dtd requires a zvol sub-tag, but I think if no_swap is
true that doesn't make sense!)
similar for a dump zvol.
>
> Currently if swap/dump is not specified it will be created by default
> (if there is enough space).
This only applies to a zvol based swap device.
The swap *slice* will only be created in a low-memory situation, but by adding
the swap attribute to a <slice> tag. then we are allowing the user to make a
decision on this themselves if it's more appropriate to a zvol swap.
>
> In pre-cud schema, setting size to "0" would result in swap not being
> created.
In the current target.dtd, you use the no_swap attribute as above.
>
> In current cud target.dtd there is a no_swap and no_swap attribute which
> defaults to False, however if set to true indicates to not create
> swap/dump, something like this needs to be maintained here.
>
Agreed, it should be maintained, but I was only talking about swap *slices*
before, not swap zvols...
I believe we should include the existing swap/dump tags in the target.dtd in the
<logical> section alongside <zpool> tag.
Although, to be consistent with what we've done with disks and zpool/vdev, I
think that this looks wrong due to duplication of <zvol> tags:
<target>
<disk>
<disk_name name="c3t0d0" name_type="ctd" zpool="rpool"/>
</disk>
<logical>
<zpool name="rpool" action="create" is_root="true" />
<dataset>
<zvol name="rpool/swap_zvol" size="12G" action="create"/>
<zvol name="rpool/dump_zvol" size="12G" action="create"/>
</dataset>
</zpool>
<swap>
<zvol name="rpool/swap_zvol" size="12G" action="create"/>
</swap>
<dump>
<zvol name="rpool/dump_zvol" size="12G" action="create"/>
</dump>
</logical>
</target>
And would be more correct as:
<target>
<disk>
<disk_name name="c3t0d0" name_type="ctd" zpool="rpool"/>
</disk>
<logical>
<zpool name="rpool" action="create" is_root="true" />
<dataset>
<zvol name="swap_zvol" size="12G" action="create" use="swap"/>
<zvol name="dump_zvol" size="12G" action="create" use="dump"/>
</dataset>
</zpool>
</logical>
</target>
'use' could be one of : none|swap|dump
Also, it would be invalid to specify a:
<zvol ... use=swap/>
when you later had a:
<swap no_swap=true/>
Thanks,
Darren.
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
