some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

[Christoph Anton Mitterer]

[-- Attachment #1: Type: text/plain, Size: 1647 bytes --]

Hi.

Christian Kujau suggested in the wiki[] to post project ideas to the
list to give them some possible wider discussion.

So far I've had these ideas:

1) NFS 4 ACLs[1]
Not sure whether it has been proposed and/or rejected before),... but it
would be nice if it was a goal for btrfs to support NFS4 ACLs.

These are much more powerful than POSIX.1e ACLs... and I think it would
be nice if all the major Linux filesystems would support them in
addition or as an alternative to the POSIX.1e ACLs. Of course one would
have to think about a schema what happens when there are ACLs of both
types in place for a file, whether this is allowed at all or not and how
rules would be merged.


2) DUP of data
I think btrfs already supports to have the same blocks copied on
multiple locations on the same device for resilience,... but only for
meta-data...
I think ZFS supports that for data as well...
Advanced code could try to spread such replicas as far as possible (i.e.
different spindles on rotating drives, ... different chips on flash
based storage)


3) allow to tell btrfs which devices are really distinct in a RAID
For some reasons people may have setups e.g. like this:
/dev/sda1
/dev/sda2
/dev/sdb1
/dev/sdc1
Now one could create a RAID5 on them, but apparently sda1 and sda2 are
on the same disk... so I'd like a way to tell btrfs that these are the
"same" and not really that redundant.


Cheers,
Chris.

[0] https://btrfs.wiki.kernel.org/index.php?title=Talk%3AProject_ideas&action=historysubmit&diff=26841&oldid=26831
[1] https://btrfs.wiki.kernel.org/index.php/Talk:Project_ideas#NFS4_ACLS

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 5165 bytes --]

Re: some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

[Goffredo Baroncelli]

On 06/02/2014 07:44 PM, Christoph Anton Mitterer wrote:
> Hi.
> 
> Christian Kujau suggested in the wiki[] to post project ideas to the
> list to give them some possible wider discussion.
> 
> So far I've had these ideas:
> 
> 1) NFS 4 ACLs[1]
> Not sure whether it has been proposed and/or rejected before),... but it
> would be nice if it was a goal for btrfs to support NFS4 ACLs.
> 
> These are much more powerful than POSIX.1e ACLs... and I think it would
> be nice if all the major Linux filesystems would support them in
> addition or as an alternative to the POSIX.1e ACLs. Of course one would
> have to think about a schema what happens when there are ACLs of both
> types in place for a file, whether this is allowed at all or not and how
> rules would be merged.

There is (was ?) a project to address that: richacl http://www.bestbits.at/richacl/.
This is not a btrfs project, but a linux kernel project because from a filesystem  POV the implementation requires to store some information in a extended attributes: the check is performed at the VFS level.

[...]
> 
> 
> 3) allow to tell btrfs which devices are really distinct in a RAID
> For some reasons people may have setups e.g. like this:
> /dev/sda1
> /dev/sda2
> /dev/sdb1
> /dev/sdc1
> Now one could create a RAID5 on them, but apparently sda1 and sda2 are
> on the same disk... so I'd like a way to tell btrfs that these are the
> "same" and not really that redundant.

Hugo Mill worked on something related to that [1]

[1] "Thoughts on RAID nomenclature" http://www.spinics.net/lists/linux-btrfs/msg33782.html

> 
> 
> Cheers,
> Chris.
> 
> [0] https://btrfs.wiki.kernel.org/index.php?title=Talk%3AProject_ideas&action=historysubmit&diff=26841&oldid=26831
> [1] https://btrfs.wiki.kernel.org/index.php/Talk:Project_ideas#NFS4_ACLS
> 


-- 
gpg @keyserver.linux.it: Goffredo Baroncelli (kreijackATinwind.it>
Key fingerprint BBF5 1610 0B64 DAC6 5F7D  17B2 0EDA 9B37 8B82 E0B5

Re: some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

[Christoph Anton Mitterer]

[-- Attachment #1: Type: text/plain, Size: 914 bytes --]

On Tue, 2014-06-03 at 19:03 +0200, Goffredo Baroncelli wrote: 
> There is (was ?) a project to address that: richacl http://www.bestbits.at/richacl/.
> This is not a btrfs project, but a linux kernel project because from a filesystem  POV the implementation requires to store some information in a extended attributes: the check is performed at the VFS level.

> > Now one could create a RAID5 on them, but apparently sda1 and sda2 are
> > on the same disk... so I'd like a way to tell btrfs that these are the
> > "same" and not really that redundant.
> 
> Hugo Mill worked on something related to that [1]
> 
> [1] "Thoughts on RAID nomenclature" http://www.spinics.net/lists/linux-btrfs/msg33782.html


Thanks for the info... should I put those 3 ideas to the
https://btrfs.wiki.kernel.org/index.php/Project_ideas page? Or is that
just meant to be edited by the developers?


Cheers,
Chris.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 5165 bytes --]

Re: some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

[Goffredo Baroncelli]

On 06/07/2014 03:34 AM, Christoph Anton Mitterer wrote:
> On Tue, 2014-06-03 at 19:03 +0200, Goffredo Baroncelli wrote:
>> There is (was ?) a project to address that: richacl
>> http://www.bestbits.at/richacl/. This is not a btrfs project, but a
>> linux kernel project because from a filesystem  POV the
>> implementation requires to store some information in a extended
>> attributes: the check is performed at the VFS level.
> 
>>> Now one could create a RAID5 on them, but apparently sda1 and
>>> sda2 are on the same disk... so I'd like a way to tell btrfs that
>>> these are the "same" and not really that redundant.
>> 
>> Hugo Mill worked on something related to that [1]
>> 
>> [1] "Thoughts on RAID nomenclature"
>> http://www.spinics.net/lists/linux-btrfs/msg33782.html
> 
> 
> Thanks for the info... should I put those 3 ideas to the 
> https://btrfs.wiki.kernel.org/index.php/Project_ideas page? Or is
> that just meant to be edited by the developers?
> 

You can add new ideas to the wiki pages, supporting by link and other info were available. This is the real nature of the wiki pages.

However I am still convinced richacl/nfs4 acl are not strictly related to a specific filesystem.

BR

> Cheers, Chris.
> 


-- 
gpg @keyserver.linux.it: Goffredo Baroncelli (kreijackATinwind.it>
Key fingerprint BBF5 1610 0B64 DAC6 5F7D  17B2 0EDA 9B37 8B82 E0B5

Re: some project ideas: NFS4 ACLs, resilience on the same device, allowing to specify which devices are "distinct" in a RAID

[Christoph Anton Mitterer]

[-- Attachment #1: Type: text/plain, Size: 412 bytes --]

On Sat, 2014-06-07 at 10:25 +0200, Goffredo Baroncelli wrote: 
> You can add new ideas to the wiki pages, supporting by link and other info were available. This is the real nature of the wiki pages.
I've added some stuff now:
https://btrfs.wiki.kernel.org/index.php?title=Project_ideas&action=historysubmit&diff=27001&oldid=26611

@upstream-devs: Just drop the ideas wich don't fit :)


Cheers,
Chris.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 5313 bytes --]