Andreas Petzold wrote:
>         Hi,
> 
> on SL4.4 x86_64 I've been trying to figure out why some of my machines
> correctly pull in the new kernel-module-openafs-<new kernel version> when the
> kernel is upgraded to a new version and some other machines don't do it
> correctly.
> 
> I looked at the python source of the kernel-modules yum plugin which comes
> with the yum rpm. The plugin is using the 'provides' part of the rpm header
> of all installed kernel-module-XXXX type packages to determine if a new
> version of the rpm is available and needs to be installed.
> 
> So I checked a few packages:
> 
> $ rpm -q --provides kernel-module-openafs-2.6.9-42.0.2.ELsmp
> kernel-module-openafs = 0:1.4.1-0.11.SL
> openafs-kernel
> kernel-module-openafs-2.6.9-42.0.2.ELsmp = 0:1.4.1-0.11.SL
> 
> $ rpm -q --provides kernel-module-openafs-2.6.9-67.0.4.ELsmp
> openafs-kernel
> kernel-module-openafs-2.6.9-67.0.4.ELsmp = 0:1.4.4-46.SL4
> 
> And now I do understand what's going wrong. The kernel-module plugin requires
> the line "kernel-module-openafs" to be present from which it constructs the
> new package name kernel-module-openafs-<new kernel>. However this line in
> the 'provides' part of the rpm header is not present in the more recent rpms.
> On some of my machines I've cleaned up very old versions of the
> kernel-module-openafs packages as they are not automatically removed even
> when the corresponding kernel is removed and on these machines the update
> doesn't work correctly. On the machines that still have the old packages the
> update works fine.
> 
> I haven't checked if this problem also occurs with other kernel module
> packages.
> 
> So my question is, is this a bug in the packaging of the kernel-module-XXX
> type packages or is this a bug in the kernel-modules yum plugin?
> 
>         Cheers,
> 
>                 Andreas

Hi Andreas,
It is probrubly the yum plugin.  And it is possible that some machines have one 
plugin, and others another.  SL 4.6 started having a newer yum plugin.  It 
works better.
Why was this not pushed out as an errata, and hense, to all SL4 machines?
It updated the yum, slightly changing it's behavior (I can't remember which 
behaviour, it was something that rarely happens I believe).
And nobody was complaining.
Try using the yum, and the kernel-module plugin, from SL 4.6, and see if this 
gives you the behaviour you are looking for.
Troy
-- 
__________________________________________________
Troy Dawson  [log in to unmask]  (630)840-6468
Fermilab  ComputingDivision/LCSI/CSI DSS Group
__________________________________________________