[Spacewalk-list] Package collision questions

[jeff macfarland]

I've run into this more than a couple times. Two fixes I am aware of

https://bugzilla.redhat.com/show_bug.cgi?id=1076490
https://bugzilla.redhat.com/show_bug.cgi?id=1287829

I submitted the latter but haven't tried out a nightly build so I am just
waiting for the next release to see if it fixed my particular issue. In the
meantime, I am resigned to removing an offending package manually and
filtering out from the sync (typically on an updates channel). This works
for me because I don't care which package is used. For the conflicting
packages I've seen (all from centos.org), all are the same minus a
different signing date.


On Fri, Jan 22, 2016 at 11:25 AM, Kalchik, Jeffery <JDKalchik at landolakes.com
> wrote:

> Good morning, all.
>
>
>
> I have a long standing situation that isn’t showing any signs of
> improvement.  In short, I have a number of situations where a particular
> package name exists with multiple checksums & backend .rpm files.
>
>
>
> Some background:
>
>
>
> I have 5 major release trees (base channel and child channels) in
> Spacewalk (2.4,)  CentOS6, CentOS7, Oracle Linux 5, Oracle Linux 6, and
> Oracle Linux 7.  The release channels have all been created through
> spacewalk-common-channels.  I also have some extra child channels in each
> tree for things like a local utilities channel and application channels.
> Each release tree is cloned into development, QA, & production trees.  All
> client systems are registered to a dev, QA or production tree, never to a
> release tree/channel.
>
>
>
> A package may exist in multiple channels in a tree, downloaded from
> separate repositories.  This can cause problems during a kickstart, when a
> package with a different checksum gets sent down to anaconda.  To make
> things worse, the kickstart problem does not always occur.   I also suspect
> that it could be an issue in spacecmd, as commands like ‘package_remove
> PKG’ don’t allow me to specify which package to remove.  Yes,
> ‘softwarechannel_removepackage PKG’ helps, I do limit each channel to a
> single repository, there shouldn’t be any collisions within a given channel.
>
>
>
> Here’s an example.  I’ve written a Perl script to spin through the base
> channels, generate the channel list for that tree, & find all packages with
> multiple IDs and checksum.
>
>
>
> librepo-1.7.16-1.el7.x86_64
>
> 128863
> centos7-x86_64,dev-centos7-x86_64,prod-centos7-x86_64,qa-centos7-x86_64
>
> 136116
> dev-epel7-centos7-x86_64,dev-epel7-oraclelinux7-x86_64,epel7-centos7-x86_64,epel7-oraclelinux7-x86_64,prod-epel7-centos7-x86_64,prod-epel7-oraclelinux7-x86_64,qa-epel7-centos7-x86_64,qa-epel7-oraclelinux7-x86_64
>
>
>
> Rather obviously, this output lists all channels where that particular
> package ID exists.  (might be a good script enhancement to limit the
> channels to only this particular tree.)
>
>
>
> Package 128863 has been downloaded from the CentOS7 repository.  Package
> 136116 has been downloaded from the Fedora Project’s Extra Packages for
> Enterprise Linux.  Both packages should be perfectly valid, but built by 2
> different organizations and due to different build environments, have
> different checksums.
>
>
>
> Is this an issue for anyone else?  How have you addressed this?
>
>
>
> Jeff Kalchik
>
> Systems Engineering
>
> Land O’Lakes
>
>
>
>
> This message may contain confidential material from Land O'Lakes, Inc. (or
> its subsidiary) for the sole use of the intended recipient(s) and may not
> be reviewed, disclosed, copied, distributed or used by anyone other than
> the intended recipient(s). If you are not the intended recipient, please
> contact the sender by reply email and delete all copies of this message.
>
> _______________________________________________
> Spacewalk-list mailing list
> Spacewalk-list at redhat.com
> https://www.redhat.com/mailman/listinfo/spacewalk-list
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/spacewalk-list/attachments/20160122/b37c9f4e/attachment.htm>