On Jun 4, 2008, at 16:59, William L. Maltby wrote:
As to your specific problem, since hardware is not common among the users reporting problems, I suspect that the only commonality is the configuration *beginning* with the automated install. My thought is there is some flaw in it that becomes common to (almost?) all the installations.
Actually, the hardware is mostly the same for all users (Lenovo ThinkCentre desktop towers). And it was working fine with CentOS 4.6. The kickstart scripts are mostly the same from 4.6 with adjustments for the changes in 5.X. The one new thing that I forgot to mention is that at the same time of the upgrade we switched NIS domains. The old NIS server was an old Solaris system, and the new one is integrated with Windows (I think it's using Windows Services for Unix). But I don't think a different NIS infrastructure can cause the instability I'm seeing.
It might be instructive to do one real manual install (that matches as closely as possible the automated results) on a representative system and an automated one. Then map the differences.
I really trust kickstart, so I don't think that is the issue.
Just as general notes from astutely watching this list, I guess mixed 32/64 bit libs could be an issue. Components from mixed repositories? All yum updated (been several things fixed since original 5.1 - I don't know if they are related).
No mixed environment here. Everything is 32-bit. I am running the centosplus kernel because of the NFS performance issue.
Have you examined the logs for any machines that seem consistently problematic?
Yes, I have and found nothing interesting. I did have another gnome- terminal crash tonight, and I'll update the other thread shortly.
Alfred