On Thu, Feb 14, 2019 at 12:51 PM Saul Wold <sgw@linux.intel.com> wrote:

Folks,

I was doing some experimentation with an un-patched CentOS and running
config_controller. One of the main issues I found is that doing the
initial installation and execution discovered many un-resolved runtime
requirements.

Thanks for looking into this Saul, I think this is a good thing to do to work towards getting a understanding of dependencies.
 

I will start sending some pull requests to fault, metal, and config with
more detailed "Requires:" statements.

Another item is that since that we are rebuilding openstack-keystone
among other openstack related packages with additional configuration and
scripts, which are needed for controller-0. In the stx-integ (base OS)
case, we re-factored many of the packages to remove configuration and
additional scripts to a separate package, I would like to see something
similar here for packages are are needed for controller-0 (ie the things
we are not installing from PyPi directly).

Do we install things directly from PyPi? When does that happen?


What I saw is that we include the CentOS-Openstack RPM repo along with,
of course, our StarlingX RPM repo. Why can't we use the CentOS-Openstack
packages directly along with some StarlingX specific additions in a
seperate package, rather than creating a new package with both upstream
and StarlingX content.


I don't know what the extra things are that we are packaging, but if they are only helper scripts and the like and don't affect the actual keystone code then I'd hope we would use the upstream RPMs.

My two cents. :)

Thanks,
Curtis

 
Thoughts,

Sau!





_______________________________________________
Starlingx-discuss mailing list
Starlingx-discuss@lists.starlingx.io
http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss


--
Blog: serverascode.com