Lustre discuss - Apr 2009 - Lustre 1.8 patched kernel for RHEL5 clients?

Is there ANY reason for running a patched lustre 1.8 RHEL5 kernel vs.  
an unpatched RHEL5 kernel on a client machine? My impression so far is  
that there is no need (assuming that the unpatched kernel is newer  
than 2.6.???)  Just wanted to confirm that there is no slight  
performance advantage for RHEL5 lustre clients that run a patched  
kernel.

------
Yemi

On Apr 17, 2009  06:06 -0700, Adeyemi Adesanya wrote:
> Is there ANY reason for running a patched lustre 1.8 RHEL5 kernel vs.  
> an unpatched RHEL5 kernel on a client machine? My impression so far is  
> that there is no need (assuming that the unpatched kernel is newer  
> than 2.6.???)  Just wanted to confirm that there is no slight  
> performance advantage for RHEL5 lustre clients that run a patched  
> kernel.
No, there isn''t really a "patched" client for RHEL5 - there
is only
an unpatched client.  The reason there is a separate kernel is for
servers, which are patched, and we want to allow clients mounts
on these kernels so there is a duplicate set of modules built for
the patched server kernel.  The client code used is identical.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.

Hello,

for clients, main difference is two patches -
first is adjust stack size to 8k, this need for run nfs and lustre
clients at same node, otherwise can be stack overflow.
second patch is light changes in dcache - which allow less tricky work
with dcache.

any big performance changes should be don''t exist. 

On Fri, 2009-04-17 at 06:06 -0700, Adeyemi Adesanya
wrote:
> Is there ANY reason for running a patched lustre 1.8 RHEL5 kernel vs.  
> an unpatched RHEL5 kernel on a client machine? My impression so far is  
> that there is no need (assuming that the unpatched kernel is newer  
> than 2.6.???)  Just wanted to confirm that there is no slight  
> performance advantage for RHEL5 lustre clients that run a patched  
> kernel.
> 
> ------
> Yemi
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at lists.lustre.org
> http://lists.lustre.org/mailman/listinfo/lustre-discuss
-- 
Alex Lyashkov <alexey.lyashkov at sun.com>
Lustre Group, Sun Microsystems

On Saturday 18 April 2009, Alex Lyashkov wrote:
> Hello,
>
> for clients, main difference is two patches -
> first is adjust stack size to 8k, this need for run nfs and lustre
> clients at same node, otherwise can be stack overflow.
A late comment. This only applies to 32-bit RHEL since x86_64 already uses 8K 
kernel stack.

/Peter
> second patch is light changes in dcache - which allow less tricky work
> with dcache.
>
> any big performance changes should be don''t exist.
>
> On Fri, 2009-04-17 at 06:06 -0700, Adeyemi Adesanya wrote:
> > Is there ANY reason for running a patched lustre 1.8 RHEL5 kernel vs.
> > an unpatched RHEL5 kernel on a client machine? My impression so far is
> > that there is no need (assuming that the unpatched kernel is newer
> > than 2.6.???)  Just wanted to confirm that there is no slight
> > performance advantage for RHEL5 lustre clients that run a patched
> > kernel.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
Url :
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090507/ec338213/attachment.bin