Chris Kirby wrote:
> Darren J Moffat wrote:
>> Mark Maybee wrote:
>>
>>> +    refreserv=size | none
>>
>>
>> Do you really need to abbreviate this, I had to read this three times 
>> before I could work out what it was (I kept thinking serve what ?). 
>> Maybe "refreserved" instead (ie create not creat :-)).
> 
> How about "refreservation", which matches the current "reservation"
> property and the "refquota"/"quota" scheme?

Seems fine to me.

> 
>>
>> I'm also having a hard time working out what the interaction between 
>> the existing "quota" and "reserveration" properties are and these 
>> ones.  I don't think the text in the man page is sufficient to allow 
>> people to work out what combinations of quota/refquota they need to 
>> set.   Maybe some more examples would help.
>>
>> Is there a case where you would want to set both quota and refquota ? 
>> If not would it make sense to have a "quotatype" property instead that 
>> determined if the quota included snapshots and descendents.
> 
> Setting both quota and refquota would be useful in a university
> environment, where refquota is what limits a user's active dataset,
> and the (presumably larger) quota limits the dataset+snapshots.
> 
> One way to think of this is that quota/reservation are for sysadmins
> and refquota/refreservation are for end users.

Could you provide an example please.

Is it an error to set refquota higher than quota ?

>> I know that "referenced" is the correct ZFS term but I'm not sure it 
>> is a term that users/admins can easily relate to.  I can't at the 
>> moment come up with a better naming scheme though - so unless anyone 
>> else can I'm okay for this case to be approved with the current naming.
>>
> 
> We wanted to link the naming to the "referenced" property, since
> that's what shows up in "zfs list" and "zfs get" output, and
> the byte count comparisions are done against that value.

Okay fair enough.

-- 
Darren J Moffat

Reply via email to