[elrepo] Announcement: EL7 Updated kmod-nvidia package for RHEL7.6

[Phil Perry]

On 07/11/2018 22:07, Rejean Lepage wrote:
> 
> There is a good chance I am wrong and I am sorry for that
> (my first language is french-canadian so sorry if something does not 
> make sense)
> but I really think there is something wrong.
> 

Yes, there is something wrong. The thing that is wrong is that CentOS 
and SL have yet to release their clones of RHEL7.6. As soon as they do 
the issues you are experiencing will just disappear.

> 1)
> for every version SL or CentOS systems (that are not yet 76) having 
> elrepo enabled will cause a failure if we try to install or update 
> nvidia-340xx
> 
> % yum --enablerepo=elrepo install kmod-nvidia-340xx
> 
> elrepo | 2.9 kB  00:00:00
> Resolving Dependencies
> --> Running transaction check
> ---> Package kmod-nvidia-340xx.x86_64 0:340.107-2.el7_6.elrepo will be 
> installed
> --> Processing Dependency: nvidia-x11-drv-340xx = 340.107 for package: 
> kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> --> Processing Dependency: kernel(sg_alloc_table_from_pages) = 
> 0xf25d238b for package: kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> --> Processing Dependency: kernel(i2c_del_adapter) = 0x28a636e9 for 
> package: kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> --> Processing Dependency: kernel(i2c_add_adapter) = 0xa6f8da46 for 
> package: kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> --> Processing Dependency: kernel(drm_legacy_pci_init) = 0xc30e01c8 for 
> package: kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> --> Processing Dependency: kernel(drm_legacy_pci_exit) = 0x0a05a9f3 for 
> package: kmod-nvidia-340xx-340.107-2.el7_6.elrepo.x86_64
> ...
> 

Correct, because the Required kernel version is not yet available on 
CentOS/SL. As soon as CentOS/SL catch up, the problem will disappear. 
There is nothing here for us to fix.

> 2)
> I did try to be more specific but it failed anyhow... I tried every 
> possible combination.
> 
>     % yum --disablerepo=\* --enablerepo=elrepo install
>     kmod-nvidia-340xx-340.107-1
>     % yum --disablerepo=\* --enablerepo=elrepo install
>     kmod-nvidia-340xx-340.107-1.el7_5
>     % yum --disablerepo=\* --enablerepo=elrepo install
>     kmod-nvidia-340xx-340.107-1.el7_5.elrepo.x86_64
>     % yum --disablerepo=\* --enablerepo=elrepo install
>     kmod-nvidia-340xx-340.107-1.el7_5.elrepo.x86_64.rpm
>     No package kmod-nvidia-340xx-340.107-1.el7_5.elrepo.x86_64.rpm
>     available.
>     Error: Nothing to do
> 
>     then *you saying this made me wonder* ... so I cleaned yum cache and
>     it worked without the .rpm (which still fail)
>          yum clean all
>          yum --disablerepo=\* --enablerepo=elrepo install
>     kmod-nvidia-340xx-340.107-1.el7_5.elrepo.x86_64
> 
>     great one thing solved. next,
> 

Great :-)

> 3)
>> Alternatively, you can add a temporary wildcard exclude for all el7.6 
>> elrepo packages until you have el7.6 available to install:
>>
>> exclude=*el7_6.elrepo* 
> 
> exclude=*el7_6.elrepo*       yes does work for now.
> 
> Although it does work it require us to reconfigure several hundred of 
> systems and change a repo config file on all system that may be power on 
> or off.
> We also need to prevent issues for future version and in example 
> el7_[6,7,8,9]
> and instead of excluding 7*6 is there a way to include ONLY *7_5* ? (for 
> EL75 systems)
> 

We have been looking at technical solutions to this issue for a long 
time. I tried to write a yum plugin to prevent yum offering packages 
where the prerequisite kernel requirements could not be met, but 
unfortunately my python skills are not quite up to the job (we would 
welcome contributions here from anyone with python experience).

Another less elegant solution might be to ship a simple rpm package 
(something like elrepo-centos-compat) that provides a yum conf file with 
a wildcard exclude:

exclude=*el7_6.elrepo*

and then as soon as CentOS/SL release their el7.6, we update the package 
to update the exclude to anticipate the next release:

exclude=*el7_7.elrepo*

It would be then be your choice if you wanted to install that and allow 
us to manage your yum configuration, or whether you want to manage the 
configuration of your own systems.

> 4)
> I think
> Having elrepo to be directly available by major and minor version would 
> be a way to fix this and it as Nothing that I now to do with CentOS.
> Scientific Linux repos are provided to be accessible by major and minor 
> version 7.3, 7.4 , 7.5  or latest 7x version if we want.
> So a system running 7.4 or 7.5 will not update to 7.6 unless 7x repo is 
> made available on that system.
> 
> Sorry to insist but this update does fail all your CentOS and Scientific 
> Linux user.
> Releasing el7_6 update is an impossible version to us - all clone users.
> 

We have a choice to make here, in the absence of a technical solution. 
We support RHEL and compatible clones. You are only viewing the issue 
from your own perspective, as a CentOS user. If we were to cater for 
CentOS users and not release 7.6 updates until CentOS have released 
their version of el7.6 (maybe a month later), then users who are running 
genuine RHEL would really have a broken system. In the case of nvidia 
packages, once they update to rhel7.6, their graphical systems would not 
boot X and they would be left with a broken system. This is very 
different to the issues you are experiencing.

You, on the other hand, simply have a few warning messages in yum that 
are relatively easy to silence if you can not bear to see them for a few 
weeks whilst CentOS catches up. All the warning messages are telling you 
is that CentOS has yet to catch up, something you knew anyway, so I 
really do not understand why you are so upset about that. Your system is 
not 'broken' in the same way as the RHEL user's system would be broken 
above.

> 5)
>> The way it is configured now *PREVENTS* you from breaking your system.
> 
> I have been using elrepo for years but I am new on the list ...
> I do fail to see the new way that *now PREVENT* from braking systems...
> To me the way it is configured now does brake systems because EL7_6 is 
> made available to all EL7.4 and EL7.5 systems.
> errors are  liste in 1).
> 
> What is supposed to be updated that is not ?
> 
> Please tell me what am I missing here.
> 

EL7.4 and EL7.5 systems (or any point release) are simply points in time 
and are not supported configurations. Elrepo supports el6 and el7. It is 
only possible to support the latest release. It is simply not possible 
for us to support any number of previous configurations from some point 
in time.

What we need to do is find a technical solution that makes it easier for 
you to manage the period where you are not running the very latest 
release because your distro is lagging behind RHEL.