Quoting Martin Hamant mh@accelance.fr:

on my freshly updated system ( 3.3 ):

# cat /etc/redhat-release CentOS release 3.3 (final) # rpm -qa | wc -l 483

And no such error message ...

Indeed, in fact I'd be surprised if everyone was having this problem; I don't suppose 3.3 would have been released if they were.

What I am suggesting is that this is a problem with my software config (rather than hardware). Everything on the two machines this has happened on is rpm'd pretty much all of it sourced from centos or dag.

I do have a lot of software installed, and I'm going to try removing some of it (the bits not sourced from centos/dag) to try and pin-point the problem.

...

Nothing happened to the machine whilst it was updating, no errors appeared whilst yum was running.

After yum has finished there are the following files in /var/lib/rpm:

__db.001 __db.002 __db.003

I have these files too.

Yeah, the only reason I deleted them is because googling suggested it was worth doing before a rebuilddb. On the first machine I did this I didn't delete those files and the rebuilddb didn't help.

...

Anybody else have a better solution?

Are you sure the MD5 of your CentOS's file is ok ? Maybe a corrupt Rpm or whatever ... i don't know.

Unfortunately, I haven't turned off yum's gpg checking so that doesn't seem very likely.

Tom.