zfs discuss - May 2008 - ZFS and disk usage management?

After struggling for some time to try and wedge a ZFS file server into
our environment, I have come to the conclusion that I''m simply going to
have to live without quotas. They have been immensely useful in the past
5 years or so in allowing us to keep track of which groups are hogging
disk space, and even finding a bug in one manufacturing/engineering tool
which occasionally crashed in a way which generated 4GB files when it
did.
 
The problem is the fact that NFS mounts cannot be done across
filesystems as implemented with ZFS and Solaris 10. For example, we have
client machines mounting to /groups/accounting... but we also have
clients mounting to /groups directly.
 
I know the zfs answer/dogma is "automounts", but it''s not
that simple. I
have no good way to know what is being mounted in which manner (blame
the NAS 5320''s boatload of bugs there... I could go on and on there in
a
curse-filled tirade), so the only real way to know is to migrate and
find out what breaks. Not good. Furthermore, there is no reasonable
"fix", anyway, other than some serious automount voodoo.
 
So, this means making the zfs filesystems at the /groups level instead
of the /groups/accounting level as I had expected to do... meaning we
can''t implement quotas in any reasonable manner that I know of.
 
That given, so I have any good options for monitoring usage of
subdirectories within my ZFS filesystems without going through a "du -sh
/groups/*" every night? It sure seems like a kludge.
 
thanks
 
johnS
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20080505/0f398fde/attachment.html>

On Mon, 5 May 2008, John.Stewart at emerson.com wrote:
> The problem is the fact that NFS mounts cannot be done across
> filesystems as implemented with ZFS and Solaris 10. For example, we have
> client machines mounting to /groups/accounting... but we also have
> clients mounting to /groups directly.
On my system I have a /home filesystem, and then I have additional 
logical-per user filesystems underneath.  I know that I can mount 
/home directly but I currently automount the per-user filesystems 
since otherwise user permissions and filesystem quotas are not visible 
to the client for anything other than Solaris 10.

I assume that ZFS quotas are enforced even if the current size and 
space free is not included in the user visible ''df''.  Is that
not
true?

Presumably applications get some unexpected error when the quota limit 
is hit since the client OS does not know the real amount of space 
free.

Bob
=====================================Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

> I assume that ZFS quotas are enforced even if the current 
> size and space free is not included in the user visible
''df''.
>  Is that not true?
> 
> Presumably applications get some unexpected error when the 
> quota limit is hit since the client OS does not know the real 
> amount of space free.
In my experience, I simply couldn''t implement Solaris-level quotas at
all for ZFS filesystems.

johnS

John.Stewart at emerson.com wrote:
>> I assume that ZFS quotas are enforced even if the current 
>> size and space free is not included in the user visible
''df''.
>>  Is that not true?
>>
>> Presumably applications get some unexpected error when the 
>> quota limit is hit since the client OS does not know the real 
>> amount of space free.
>>     
>
> In my experience, I simply couldn''t implement Solaris-level quotas
at
> all for ZFS filesystems.
>
>   
That''s my understanding also. I''m not clear (but I think I can
guess) on
the exact difference between reservations, and quotas but from what I 
understand ZFS implements it''s own ''Tree Quotas'' ,
that limit the space
consumed by a directory and everything below it. It does not 
(currently?) support a tradtional unix "User/Group Quotas", where the 
space consumed by files owned by a user or group are limited no matter 
where in the file system the are located.

I don''t think User/Group quotas have been ruled out, or are technically
not feasible. If I recallcorrectly, I think they have just been left on 
the low priority list since the consensus of the developers is that Tree 
quotas are enough for now, especially if you follow the advice to create 
a ZFS for each user. This works fine for Home Dirs since for the most 
part only one user will be writing below that directory. However this 
doesn''t really help where you want multiple users or groups to write to
the same dir. tree, and want to stop the users or groups from consuming 
all the space and blocking the others from using the share space.

   -Kyle


> johnS
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>

Kyle McDonald wrote:
> John.Stewart at emerson.com wrote:
>>> I assume that ZFS quotas are enforced even if the current 
>>> size and space free is not included in the user visible
''df''.
>>>  Is that not true?
>>>
>>> Presumably applications get some unexpected error when the 
>>> quota limit is hit since the client OS does not know the real 
>>> amount of space free.
>>>     
>> In my experience, I simply couldn''t implement Solaris-level
quotas at
>> all for ZFS filesystems.
>>
>>   
> That''s my understanding also. I''m not clear (but I think
I can guess) on
> the exact difference between reservations, and quotas but from what I 
> understand ZFS implements it''s own ''Tree Quotas''
, that limit the space
> consumed by a directory and everything below it. It does not 
> (currently?) support a tradtional unix "User/Group Quotas", where
the
> space consumed by files owned by a user or group are limited no matter 
> where in the file system the are located.
Here''s a good description of ZFS quotas and reservations:

http://blogs.sun.com/markm/category/ZFS

We have since added refquota and refreservation, which do not include
the space consumed by snapshots.

-Chris