[lustre-discuss] pool restrictions

[Crowe, Tom]

Thanks Dennis,

I failed to include a very important piece if information. This is on a
2.7.0 filesystem.

# cat /proc/fs/lustre/version
lustre: 2.7.0
kernel: patchless_client
build:  2.7.0-RC4--PRISTINE-2.6.32-504.8.1.el6_lustre.x86_64


I believe the patch “should” have been included.

Thanks,
Tom

On 12/18/15, 3:02 PM, "Dennis Nelson" <dnelson at ddn.com> wrote:

>See this bug.
>
>https://jira.hpdd.intel.com/browse/LU-5916
>
>-- 
>Dennis Nelson
>Mobile: 817-233-6116
>
>Applications Support Engineer
>DataDirect Networks, Inc.
>dnelson at ddn.com
>
>
>
>
>
>
>On 12/18/15, 1:52 PM, "lustre-discuss on behalf of Crowe, Tom"
><lustre-discuss-bounces at lists.lustre.org on behalf of thcrowe at iu.edu>
>wrote:
>
>>Greetings All,
>>
>>
>>I have been testing lustre pools, attempting to ³restrict² data placement
>>by setting the striping information for a top level directory of a
>>specific user/project with Œlfs setstripe ­p fsname.poolname
>>top-level-dir¹.
>>
>>This works as desired, however, when I become the non-root user that has
>>access to this top level directory, and perform a Œlfs setstripe ‹c 4¹ on
>>a new file and or directory under the top level directory, the pool
>>information is not ³inherited². The lfs setstripe command happily
>>completes, and sets the file/dir striping to 4 OST¹s, which are not
>>members of the pool.
>>
>>I believe a client side mount option of nouser_xattr ³might² prevent this
>>behavior, but it would also likely prevent ANY setstripe operations from
>>occurring on that client, as well as other extended attributes (ACL¹s,
>>etc)
>>
>>So my questions are 2 fold.
>>
>>1. Am I missing anything about how pool¹s behave? Should a file/dir¹s
>>previously defined pool be ³inherited² if no NEW pool is passed as a part
>>of the lfs setstripe command? Is there another way to ³contain² a
>>specific
>>user¹s data into a limited set of OST¹s?
>>
>>2. Is there any other way other than the client mount option nouser_xattr
>>to prevent users from altering striping? In a scenario where we are not
>>in
>>control of the clients, it would be most helpful to set a global (MDT)
>>parameter to prevent user¹s ability to use/set extended attributes.
>>
>>
>>Thanks for your input.
>>
>>-Tom Crowe
>>
>>
>>_______________________________________________
>>lustre-discuss mailing list
>>lustre-discuss at lists.lustre.org
>>http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
>