We're trying to set up an internal network of Oracle Linux yum servers. We have one master server for the company that retrieves the packages directly from Oracle, and we want to have additional servers that reposync the packages from the master server and act as local yum servers in their data centers. All the servers are virtual, running on VMware, and x86_64 architecture. We support applications that require 32-bit libraries, so we need to host them on each of the servers, too.
So far, the 64-bit packages are being distributed as expected. However, the 32-bit packages, which are present on the master server, aren't making it to the local server, just one in our proof-of-concept so far.
reposync -r reponame -p /path/to/x86_64/repo
downloads all the x86_64 packages to the specified location. However:
reposync -r reponame --arch=i386 -p /path/to/i386/repo
just extracts all the "noarch" packages from the x86_64 repository on the master server and dumps them in the i386 repository on the local server along with a couple of i386 RPMs. The reposync man page says -a ARCH or --arch=ARCH should make reposync act as if running the specified architecture, but what's happening doesn't match the behavior when reposync is left to the default/current architecture.
So: does anyone know how to get reposync's -a or --arch option to behave as the man page says it does? Thanks in advance!