RHEL5.5 kickstart procedure unable to download RPMs

[Hugh Brown]

On 03/22/2011 01:37 PM, Michael Mayer wrote:
> Hi all,
>
> i have a weird issue with RHEL5.5 and kickstart.
>
> My current setup is:
>
> All RHEL5.5 RPMs(i.e. the whole tree) and two additional repositories
> are in a subversion repo. The repo is accessible via http to the client
> to be installed.
> The client is a virtual machine hosted on an ESX server. I have modified
> boot.iso as well to contain the kickstart file which points to the
> subversion repo via the "url" and "repo" commands. The boot.iso (about
> 10 MB in size) is attached to the virtual machine as a virtual CD-ROM.
>
> If the virtual server boots up, it reads the kickstart file and then
> proceeds to download the second-stage installer from the repository. It
> sets up networking, formats the disk as wanted, browses through the RPM
> repos for dependency resolution. Once that is finished, the screen
> appears where RPMs are going to be installed.
>
> So far so good.
>
> Now things are going odd: anaconda complains about packages being
> corrupt, broken or missing. This cannot be true here because I have
> checked md5sum of the rpm both in the repo and after downloaded via
> wget. In the SVN server access logs the respective RPM is downloaded 10
> times, each time with return code 200 which for me indicates successful
> download. If I am looking on the kickstarting server, there are no RPMs
> to be seen anywhere. If I run a wget on that kickstarted server with the
> same URL the RPM is downloaded correctly. Depending on which repos I am
> using in the kickstart file, a different RPM is thought to be corrupt.
>


This is most likely a problem with the filelists.xml.gz file in your yum 
repository.  Each package gets stored with a pkgid that is hash of the 
package (the hash depends on the checksum employed by createrepo).  If 
it is complaining about a package in one of your additional 
repositories, then I'd start by making sure that a

createrepo --update /path/to/repo

has been done recently (needs to be done after any new packages are added).

If the repodata is current, then I've seen issues like this where the 
ram was going bad.

Hugh