[Lustre-discuss] lustre 1.8.3 upgrade observations

Heiko Schröter schroete at iup.physik.uni-bremen.de
Fri Jul 30 02:37:52 PDT 2010


Hello,

1) when compiling the lustre modules for the server the ./configure script behaves a bit odd.
The --enable-server option is silently ignored when the kernel is not 100% patched.

Unfortunatly the build works for the server, but during the mount the error message claims about a missing "lustre" module which is loaded and running.
What is really missing are the ldiskfs et al module(s).

My feature request is that "./configure --enable-server" should fail when it is not able to build the server modules, instead of printing the following message:

<snip>
checking if kernel defines unshare_fs_struct()... no
checking for /usr/src/linux-2.6.27.39-0.3/include/linux/lustre_version.h... no
configure: WARNING: Unpatched kernel detected.
configure: WARNING: Lustre servers cannot be built with an unpatched kernel;
configure: WARNING: disabling server build
checking whether to enable pinger support... yes
<snap>

Used SuSE-2.6-sles11 2.6.27.39-0.3 kernel source from the lustre site.

2) tunefs.lustre fails on MDS/MDT and OSS/OST.

<snip>
tunefs.lustre --ost --fsname=lustre --mgsnode=mds1 at tcp0 --verbose --param="failover.mode=failout" /dev/sdb
checking for existing Lustre data: found CONFIGS/mountdata
Reading CONFIGS/mountdata

   Read previous values:
Target:     lustre-OST0027
Index:      39
Lustre FS:  lustre
Mount type: ldiskfs
Flags:      0x2
              (OST )
Persistent mount opts: errors=remount-ro,extents,mballoc
Parameters: failover.mode=failout mgsnode=192.168.16.122 at tcp

tunefs.lustre: cannot change the name of a registered target
tunefs.lustre: exiting with 1 (Operation not permitted)
<snap>

The purpose of tunefs.lustre is not clear from the docs. The continuation from this point of "error" is not clear.
It has been confirmed (see below) that tunefs.lustre is optional when no parameters are changed.
So we decided to just carry on with the mounting of the partitions which eventually worked as expected.

Thanks and Regards
Heiko

... sweating out 3.64 gallons during upgrade ....





Natan Rutman wrote:
Sheila, can we fix this up?  Let me know what you need from me.

On Jul 27, 2010, at 10:47 PM, Heiko Schröter wrote:

> Am Dienstag 27 Juli 2010, um 18:41:24 schrieben Sie:
> 

>> 
>> You actually don't need to run tunefs unless you are changing something.  Additionally, all parameters should be "sticky" across tunefs; you can see the current settings with 'tunefs --print', add new parameters with 'tunefs --param', and erase all old params with 'tunefs --erase'.
> 
> Could this "option" make into the docs ?
> Chapter 13.2.4.2 of lustre 1.8 OP Manual reads a bit different, well, to me at least.
> An info that tunefs.lustre is only required when changing params would be nice.
> 
> Thanks and Regards
> Heiko



More information about the lustre-discuss mailing list