Ran out of disk space during yum update

[Craig White]

On Thu, 2009-02-05 at 19:50 -0500, Sam Varshavchik wrote:
> Robert Moskowitz writes:
> 
> > In the new install I did, I was not alert and did a complete yum update, 
> > and my / partition ran out.
> > 
> > 200 of 300+ packages were updated/installed, of course none cleaned.
> > 
> > Can I rescue this install by doing a yum clean all and then again do the 
> > yum update for the remaining 100+ packages?
> 
> Nope. "yum clean all" purges internal yum metadata only. You've ran into a 
> known, long-time rpm design defect. If your rpm update operation fails, 
> you'll end up with all the updated packages installed, but none of the old 
> packages removed. I've bitched about this before, I maintain that this is a 
> design defect or a bug in rpm that should be fixed, but nobody cares.
> 
> Been there, brought back the trophy. The only way to fix this is to manually 
> assemble a list of packages that should've been removed, but haven't, and 
> remove them yourself. rpm -q -a --queryformat '%{NAME}\n' returns a list of 
> all packages. By sorting them, and with some shell scripting-fu, you'll end 
> up with a list of packages names that are installed more than once -- the 
> old and the new package. You'll have to prune the list -- some packages, 
> like rpm-gpgkey, and kernel, can have multiple versions legitimately 
> installed. If you're running x86_64, you may have both 32 and 64 bit 
> versions of each package legitimately installed -- you'll have to do 
> something else, then.
> 
> Then, you'll have to take that list, and for each package, obtain the 
> version/release of the old package, then feed the result to another script 
> that removes the old version of each package.
----
yum-utils package makes this so much easier...

package-cleanup --dupes | --cleandupes

can really handle most if not all of this

Craig