Questions about ceph

Questions about ceph

[Mag Gam]

Hello,

Do ceph clients need fuse module? or is the file system provided
natively thru the ceph module? Is the filesystem POSIX compliant? For
the underlying filesystem does it have to be BRTFS? We are a heavy XFS
and ext2 shop for performance.
We are very excited about this filesystem and we would like to start
using it and start contributing by benchmarking it and submitted any
bug reports to stabilize it for distributions.

Re: Questions about ceph

[Gregory Farnum]

On Mon, Sep 13, 2010 at 4:16 AM, Mag Gam <magawake@gmail.com> wrote:
> Do ceph clients need fuse module? or is the file system provided
> natively thru the ceph module?
No, clients don't need the FUSE module if you're using the in-kernel client.

> Is the filesystem POSIX compliant?
Yes! Being POSIX-compliant is one of Ceph's missions in life.

> For the underlying filesystem does it have to be BRTFS? We are a heavy XFS
> and ext2 shop for performance.
The underlying FS doesn't need to be btrfs, but it generally has
better performance and consistency guarantees while using Ceph than
other FSes will. Ceph uses some btrfs-only ioctls to control on-disk
state for guaranteed consistency, and advanced features like snapshots
will go faster with btrfs since it supports them natively (other FSes
require lots of full copies with snapshots). Remember that the FS
characteristics of something running on [Ceph on ext2] will be very
different from running on [ext2].
Sage can talk more about this, though; he does a lot of random things
to make the on-disk stuff work nicely under different configurations.

> We are very excited about this filesystem and we would like to start
> using it and start contributing by benchmarking it and submitted any
> bug reports to stabilize it for distributions.
:) Thanks for looking at Ceph!
-Greg

Re: Questions about ceph

[Mag Gam]

Great.

Well, currently we are running RHEL 5.2 and I would really like to try
ceph without fuse. Is this still possible? Are there any instructions
on how to do this?



On Mon, Sep 13, 2010 at 11:38 AM, Guy Coates <gmpc@sanger.ac.uk> wrote:
> On 13/09/10 12:16, Mag Gam wrote:
>> Hello,
>>
>
> Hi Mag,
>
> One of the nice ceph features that only work with btrfs as the backing
> filesystem is the sub-directory accounting;
>
> ls -ald on a directory will show the size of all the files contained
> within in, and getfattr -d directory will show the number of files and
> other accounting information.  Invaluable for large file-systems!
>
> Cheers,
>
> Guy
>
>
> --
> Dr. Guy Coates, Informatics System Group
> The Wellcome Trust Sanger Institute, Hinxton, Cambridge, CB10 1HH, UK
> Tel: +44 (0)1223 834244 x 6925
> Fax: +44 (0)1223 496802
>
>
> --
>  The Wellcome Trust Sanger Institute is operated by Genome Research
>  Limited, a charity registered in England with number 1021457 and a
>  company registered in England with number 2742969, whose registered
>  office is 215 Euston Road, London, NW1 2BE.
>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Questions about ceph

[Gregory Farnum]

On Mon, Sep 13, 2010 at 4:54 PM, Mag Gam <magawake@gmail.com> wrote:
> Well, currently we are running RHEL 5.2 and I would really like to try
> ceph without fuse. Is this still possible? Are there any instructions
> on how to do this?
RHEL 5.2 is running a *really* old kernel (looks like 2.6.18?); I
don't know of anybody who's tried to run the in-kernel client on
something that old. There's a backports tree at
http://git.kernel.org/?p=linux/kernel/git/sage/ceph-client-standalone.git,
but it's only good back to 2.6.28 or so.
-Greg

Re: Questions about ceph

[Sage Weil]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1719 bytes --]

> On Mon, Sep 13, 2010 at 11:38 AM, Guy Coates <gmpc@sanger.ac.uk> wrote:
> > On 13/09/10 12:16, Mag Gam wrote:
> >> Hello,
> >>
> >
> > Hi Mag,
> >
> > One of the nice ceph features that only work with btrfs as the backing
> > filesystem is the sub-directory accounting;
> >
> > ls -ald on a directory will show the size of all the files contained
> > within in, and getfattr -d directory will show the number of files and
> > other accounting information.  Invaluable for large file-systems!

(Didn't see this one come over the list, but just to clarify:)

The recursive accounting is a function of the Ceph MDS and will work 
regardless what what backend file system you're using on the OSDs (extN, 
btrfs, xfs).  The main thing that currently doesn't work properly without 
btrfs is recovery when snapshots are being used.  (It'll work most of the 
time, but there are corner cases that aren't covered.)  This is something 
we plan to fix.

sage


> >
> > Cheers,
> >
> > Guy
> >
> >
> > --
> > Dr. Guy Coates, Informatics System Group
> > The Wellcome Trust Sanger Institute, Hinxton, Cambridge, CB10 1HH, UK
> > Tel: +44 (0)1223 834244 x 6925
> > Fax: +44 (0)1223 496802
> >
> >
> > --
> >  The Wellcome Trust Sanger Institute is operated by Genome Research
> >  Limited, a charity registered in England with number 1021457 and a
> >  company registered in England with number 2742969, whose registered
> >  office is 215 Euston Road, London, NW1 2BE.
> >
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
>