[Lustre-discuss] 1.8.1.1

[Andreas Dilger]

On 2009-12-10, at 16:37, Papp Tamas wrote:
> Well, it was not working, and by the way, my guess is that it should  
> not work. I didn't write, but /usr/src/kernels/ 
> 2.6.18-128.7.1.el5_lustre.1.8.1.1-x86_64/ belongs to the official  
> kernel-lustre-devel-2.6.18-128.7.1.el5_lustre.1.8.1.1 package.
>
> Anyway, a build an own kernel and lustre b1_8 with the patches from  
> the bug 19557:
>
> Linux meta1 2.6.18-prep #1 SMP Sun Dec 6 14:40:15 CET 2009 x86_64  
> x86_64 x86_64 GNU/Linux
>
>
> I installed it on the MDS and the tw OSSs. Unofortunately no luck:
>
> Dec  9 18:41:36 node1 kernel: Call Trace:
> Dec  9 18:41:36 node1 kernel:  [<ffffffff88491b91>] :lnet:LNetMDBind 
> +0x301/0x450
> Dec  9 18:41:36 node1 kernel:  [<ffffffff8003dacd>] lock_timer_base 
> +0x1b/0x3c
> Dec  9 18:41:36 node1 kernel:  [<ffffffff8001caa7>] __mod_timer 
> +0xb0/0xbe
> Dec  9 18:41:36 node1 kernel:  [<ffffffff8006387b>] schedule_timeout 
> +0x8a/0xad
> Dec  9 18:41:36 node1 kernel:  [<ffffffff80096ff3>] process_timeout 
> +0x0/0x5
> Dec  9 18:41:36 node1 kernel:   
> [<ffffffff888234ca>] :ost:ost_brw_write+0x137a/0x23a0
> Dec  9 18:41:36 node1 kernel:   
> [<ffffffff8859f998>] :ptlrpc:ptlrpc_send_reply+0x5c8/0x5e0

This is a different issue.  The previous stack was busy in  
lustre_hash_for_each_empty().

> Should I install this patched lustre on the clients too? Or is the  
> problem something else?

This is strictly a server-side patch.  This looks like a similar  
issue, and may be fixed by one of the other patches on bug 19557.

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