> I do not think this is really a problem with yum. I do however think it > will help because the couple of times I have run yum by hand and gotten > borked downloads it was caused by the ftp server timing out and only > doing a partial download. The next time yum is run it sees the borked > file and yum barfs. Of course pruning the borked file and down loading > it again fixes the problem. I am assuming that is what is happening > when yum is run from cron. right - my goal would be to probably implement two things 1. the retrying thing 2. maybe a command "yum sanitycheck" - so it runs through and sanitychecks all the headers, throws out old ones, etc. > I agree it does sound like a hack. Would that really help though? I > thought in addition to pruning the __db files the db needed rebuilt? > > I agree this is a Red Hat/rpm problem. I know Jeff Johnson has test packages > out there for rpm. I also know he is getting a lot of heat of late for > rpm segfaulting. Hopefully he will have a good fix for rpm and this > will be a non problem for yum. Well I talked to jeff some about this - he seemed to tell me that I should indeed just rm the __db files before a write operation in yum until the NPTL stuff get settled out. If he wasn't being sarcastic it's not terribly hard to unlink those files right before the ts is run. -sv
