On Fri, Aug 25, 2006 at 08:59:09AM -0400, seth vidal wrote: > 1. is there anything common about the 1/8th? Clearly, in that they all broke. :) Other than that, not that we can discern. We thought it might have something to do with the kernel-devel update and the installonlyn plugin (ie which kernel-devel packages were installed), but I'm pretty sure now that that's a red herring. Hmm. I wonder if the ones that broke were running 2.6.9-34.0.1.EL, and the ones that didn't, 2.6.9-34.0.2.EL. I'll check into that. But even then, what about it would make rpm/yum decide to do an erase instead of upgrade? > 2. is it possible to recreate this situation but do the transaction with > rpm as much as possible? It seems to work fine every time, although we haven't tried much along those lines at this point. I'll see what more I can do. > This feels like it is a function of the openafs package being updated > but I don't know what's going on here from this. We've updated openafs dozens of times this way with no problem. It's weird. Any guess on the discrepancy between the captured yum output vs. the logs and the reality that the package got erased instead of upgraded? -- Matthew Miller mattdm@xxxxxxxxxx <http://mattdm.org/> Boston University Linux ------> <http://linux.bu.edu/>
