Lustre discuss - May 2006 - make lustre 1.4.5 from source problem

I figure out that liblustre.h defines the same struct iattr with
"ia_attr_flags" field. So I think that llite_lic.c is using the one
defined in the linux header files "fs.h" instead of using the one
coming
with liblustre.h. Is this right?
If it is the case, how can I make llite_lib.c use the right one without
modifying the files explicitly to avoid deviation from the standard
lustre files.

Thanks again for your time.

Omar.

Omar Belakhdar wrote:
>Good morning,
>First, I''m new to lustre.
>After reading most of the documentaion coming with lustre from different
>sites, I start building a Lustre 1.4.5 for a 2.6.14.3 kernel.
>I prepared all the necessary pacthes and every think was right and the
>kernel is compiling without any problem.
>
>I also made the kernel patches for ldiskfs that lustre make will need
>for compiling.
>
>My kernel configuration doesn''t have any special config. I setup
the
>Ext2/Ext3, KALLSYMS, Nopreemption configs in the kernel, etc.
>
>When I tried to make the lustre 1.4.5 using, the provide configure
>command with "configure --with-linux=/usr/src/linux-2.6.14.3", I
get an
>error related to the llite_lib.c, saying that "struct attr" (which
is
>defined in linux/fs.h) doesn''t have  a member called
"ia_attr_flags"
>that is trying to set.
>
>I''m guessing that the configure command is getting wrong
information
>from my setting.
>
>Does anyone experience such kind of problem?
>
>Any help is welcome and thanks for your time.
>
>Omar.
>_______________________________________________
>Lustre-discuss mailing list
>Lustre-discuss@lists.clusterfs.com
>https://lists.clusterfs.com/mailman/listinfo/lustre-discuss
>
>  
>

On Dec 16, 2005  12:07 +0100, Omar Belakhdar wrote:
> When I tried to make the lustre 1.4.5 using, the provide configure
> command with "configure --with-linux=/usr/src/linux-2.6.14.3", I
get an
> error related to the llite_lib.c, saying that "struct attr"
(which is
> defined in linux/fs.h) doesn''t have  a member called
"ia_attr_flags"
> that is trying to set.
This field was removed from struct iattr in the 2.6.14 kernel.  We do
not yet support that kernel.  There is also no support in 1.4.5 for
the 2.6.12 kernel.  I''d suggest running the most recent supported
kernel for 1.4.5.1 (which has an update for this) - RHEL4 2.6.9-22.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.

Hi Andreas,
Thanks for the clarifications. I suceeded to compile and run lustre with
2.6.13.4.
However, I have a HA question concerning the OSTs. Does lustre-1.4.5
support RAID-1 or not yet (as expected from the Roadmap to be for 1.8)?
If no, what happens if the failed OST is containing a  database file,
does  the whole database  be not accessible at all or one can still
write to the database new records?

Req: I didn''t consider a shared storage for redundancy.

Thanks for your time.

Omar.


Andreas Dilger wrote:
>On Dec 16, 2005  12:07 +0100, Omar Belakhdar wrote:
>  
>
>>When I tried to make the lustre 1.4.5 using, the provide configure
>>command with "configure --with-linux=/usr/src/linux-2.6.14.3",
I get an
>>error related to the llite_lib.c, saying that "struct attr"
(which is
>>defined in linux/fs.h) doesn''t have  a member called
"ia_attr_flags"
>>that is trying to set.
>>    
>>
>
>This field was removed from struct iattr in the 2.6.14 kernel.  We do
>not yet support that kernel.  There is also no support in 1.4.5 for
>the 2.6.12 kernel.  I''d suggest running the most recent supported
>kernel for 1.4.5.1 (which has an update for this) - RHEL4 2.6.9-22.
>
>Cheers, Andreas
>--
>Andreas Dilger
>Principal Software Engineer
>Cluster File Systems, Inc.
>
>_______________________________________________
>Lustre-discuss mailing list
>Lustre-discuss@lists.clusterfs.com
>https://lists.clusterfs.com/mailman/listinfo/lustre-discuss
>
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://mail.clusterfs.com/pipermail/lustre-discuss/attachments/20051220/5dfc155d/attachment.html

On Dec 20, 2005  18:35 +0100, Omar Belakhdar wrote:
> However, I have a HA question concerning the OSTs. Does lustre-1.4.5
> support RAID-1 or not yet (as expected from the Roadmap to be for 1.8)?
> If no, what happens if the failed OST is containing a  database file,
> does  the whole database  be not accessible at all or one can still
> write to the database new records?
> 
> Req: I didn''t consider a shared storage for redundancy.
If the OSTs are running in "failover" mode, that means client IO will
block until the failed OST is recovered (currently this means shared
storage and HA software to handle service failover between OST nodes).
The RAID1 redundancy at the Lustre level is not yet implemented.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.

Good morning,
First, I''m new to lustre.
After reading most of the documentaion coming with lustre from different
sites, I start building a Lustre 1.4.5 for a 2.6.14.3 kernel.
I prepared all the necessary pacthes and every think was right and the
kernel is compiling without any problem.

I also made the kernel patches for ldiskfs that lustre make will need
for compiling.

My kernel configuration doesn''t have any special config. I setup the
Ext2/Ext3, KALLSYMS, Nopreemption configs in the kernel, etc.

When I tried to make the lustre 1.4.5 using, the provide configure
command with "configure --with-linux=/usr/src/linux-2.6.14.3", I get
an
error related to the llite_lib.c, saying that "struct attr" (which is
defined in linux/fs.h) doesn''t have  a member called
"ia_attr_flags"
that is trying to set.

I''m guessing that the configure command is getting wrong information
from my setting.

Does anyone experience such kind of problem?

Any help is welcome and thanks for your time.

Omar.