Lustre discuss - May 2009 - Mounting Lustre as an ext3 partition

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi All,

 I just wanted to check if there would be problems with mounting a
Lustre filesystem as ext3 and reading/writing to it. Would there be any
corruption or other unwanted effects to the data integrity?

 If we encounter any issues with Lustre, a backup plan would be to mount
the volume as ext3 and serve the files (for example via NFS) as a
temporary solution in an emergency.

Thanks,
Nick

- --
Nick Jennings
Technical Director
Creative Motion Design
www.creativemotiondesign.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iEYEARECAAYFAkoKj0AACgkQbqosUH1Nr8eLOQCg4SCYeJTS22TyuIYoUwcLq4r3
tBYAnRvhvsqQF4YSsAJq6Fx7LSrpf9AG
=J9mS
-----END PGP SIGNATURE-----

On May 13, 2009, at 11:13 AM, Nick Jennings wrote:
> I just wanted to check if there would be problems with mounting a
> Lustre filesystem as ext3 and reading/writing to it. Would there be  
> any
> corruption or other unwanted effects to the data integrity?
>
> If we encounter any issues with Lustre, a backup plan would be to  
> mount
> the volume as ext3 and serve the files (for example via NFS) as a
> temporary solution in an emergency.
I''m not sure to understand how you want to do this, given that data and
metadata are stored on separate ext3 filesystems (MDT & OSTs).
If you mount the MDT locally as a ext3 fs and export it via NFS, you  
will
only see empty files.

Johann

On Wed, 2009-05-13 at 11:13 +0200, Nick Jennings wrote:
> 
>  I just wanted to check if there would be problems with mounting a
> Lustre filesystem as ext3 and reading/writing to it.
It probably won''t mount as ext3 as it''s likely not compatible
enough
with ext3 any more.  ext4 if you have that perhaps, but better to just
use ldiskfs which is it''s native format.
> Would there be any
> corruption or other unwanted effects to the data integrity?
Not just by mounting it.  Of course, if you start fiddling with it''s
contents, you could cause damage to the coherency of the overall Lustre
filesystem.
>  If we encounter any issues with Lustre, a backup plan would be to mount
> the volume as ext3 and serve the files (for example via NFS) as a
> temporary solution in an emergency.
No, that won''t work.  You have to remember that a Lustre filesystem
separates the metadata (filenames, attributes, etc.) from the file data
and puts the former on the MDT and the latter on OSTs.  Only Lustre can
bring these back together to form a cohesive filesystem again.

b.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url :
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090513/9faa4fd1/attachment-0001.bin

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ah yes, of course. Thanks for the reply, I guess I wasn''t thinking that
one through completely.

So if, for example, the MDS drive dies, my entire Lustre filesystem is
unrecoverable?

Brian J. Murrell wrote:
> On Wed, 2009-05-13 at 11:13 +0200, Nick Jennings wrote:
>>  I just wanted to check if there would be problems with mounting a
>> Lustre filesystem as ext3 and reading/writing to it.
> 
> It probably won''t mount as ext3 as it''s likely not
compatible enough
> with ext3 any more.  ext4 if you have that perhaps, but better to just
> use ldiskfs which is it''s native format.
> 
>> Would there be any
>> corruption or other unwanted effects to the data integrity?
> 
> Not just by mounting it.  Of course, if you start fiddling with
it''s
> contents, you could cause damage to the coherency of the overall Lustre
> filesystem.
> 
>>  If we encounter any issues with Lustre, a backup plan would be to
mount
>> the volume as ext3 and serve the files (for example via NFS) as a
>> temporary solution in an emergency.
> 
> No, that won''t work.  You have to remember that a Lustre
filesystem
> separates the metadata (filenames, attributes, etc.) from the file data
> and puts the former on the MDT and the latter on OSTs.  Only Lustre can
> bring these back together to form a cohesive filesystem again.
> 
> b.
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at lists.lustre.org
> http://lists.lustre.org/mailman/listinfo/lustre-discuss
- --
Nick Jennings
Technical Director
Creative Motion Design
www.creativemotiondesign.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iEYEARECAAYFAkoKws4ACgkQbqosUH1Nr8cv4gCg2lb5oS4BZ1hfjtf3/YKzu8z+
D8UAoMtdkOeNFLPJ8BORpJeXKfSqZcdG
=JFoM
-----END PGP SIGNATURE-----

On Wed, 2009-05-13 at 14:53 +0200, Nick Jennings wrote:
> 
> Ah yes, of course. Thanks for the reply, I guess I wasn''t thinking
that
> one through completely.
> 
> So if, for example, the MDS drive dies, my entire Lustre filesystem is
> unrecoverable?
Yes.  That is why _reliable_ storage for the MDT is paramount.  RAID 5/6
doesn''t perform a well as RAID 1 for the MDT however, so RAID 1 is our
recommendation.

Some people also back up their MDT.  There has been lots of discussion
of that on this list.  The archives will be quite fruitful if you want
to research that option as well.

b.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url :
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090513/8416ba73/attachment.bin