[Lustre-discuss] mismatching versions between kernel and client

[Brian J. Murrell]

On Mon, 2009-03-23 at 08:31 -0700, Shi Jin wrote:
> Hi there,
> 
> I am trying to install a patchless Lustre client on a CentOS 5.2 system from RPM. I downloaded the following two RPMs:
>     *  lustre-client-1.6.7-2.6.18_92.1.17.el5_lustre.1.6.7smp.i686.rpm
>     * lustre-client-modules-1.6.7-2.6.18_92.1.17.el5_lustre.1.6.7smp.i686.rpm 
> 
> However, the kernel of my system is slightly different: 2.6.18-92.e15.

You have two choices.  Either upgrade the kernel on your system to
2.6.18_92.1.17.el5 (which is the suggested route as it gets you a newer
kernel with more recent patches applied) or rebuild lustre from the
lustre-source rpm using the kernel-devel for the kernel you wish to use.

> Therefore, the installed modules are located in a different directory in /lib/modules to the system kernel. I manually created symbolic links of the fs and net directories into the system kernel directory.

You can't/shouldn't do this.

> However, I am still not able to load the lustre modules.

Right.  This is not surprising.  You really do need to match the lustre
modules to the right kernel.

> The whole problem comes from the slight difference between the versions of the kernel and the modules. I am wondering if there is a simple way to get the modules loaded without installing a new kernel.

The only other option, as I mentioned is to build the lustre modules for
the kernel you wish to use.

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-lustre.org/attachments/20090323/606b4621/attachment.pgp>