"yummification" is necessary to have the adequate files used by the RH patch analysis
I used the information & script I found here: http://www.oracle.com/technology/tech/linux/htdocs/yum-repository-setup.html
to build a yum repository at a customer's and to re-write the population procedure we provide
The client has a fraction of a terabyte devoted to patches already;
In addition they have an existing methodology which keeps it up to date.
Is it possible to leverage that?
Ideally we would like to avoid setting up a seperate repository.
I believe you can do that with the yum-arch command. see its options, I think you can specify where the information generated by the command will be stored.
the new patchkits can use the existing repositories, you need an agent on the server w/ the repos.
'yumification' is basically using the yum-arch to strip the headers off the set of rpms so when you run it a 'headers' directory is created in the directory of rpms.
w/ 7.4.1 and the patchkit you can use the network install feature to create packages that will use payloads (rpms in this case) pulled directly from the existing location at install time instead of pulling them into the depot, however the rpms must be available from the target server. (nfs/cifs, not sure about http)
i use mrepo alot at home. you can have mrepo generate the old style metadata itself.
in the /etc/mrepo.conf.d/ (or mrepo.conf) look for the 'metadata' item - you have a bunch of different options - there is one that will use the type we need, so they can just add that type to the list.
i think you will add the option like:
metadata = repomd repoview yum
'yum' should generate the header files which is what you need.
what version of mrepo are they using?
in the patchkit nsh script job, you will then just point to the repo directory via nsh - so like:
where myrepo contains a 'headers' directory after metadata generation