[WBEL-users] yum broken after upgrade

[Kirby C. Bohling]

> > 
> > Did I miss anything? Going over the mailing lists no mention of any issues 
> > so far...
> 
> I did two things.  I went to a WBEL3 system that didn't have this
> problem and did a "tar -pzcvf rpm.tar.gz `rpm -ql rpm | grep -v '^/var'"
> and restored the files on the affected system.  No luck.
> 
> Then I rebooted for another reason -- and it worked happily from then
> on.
> 
> WTF?  Rebooting to make a database readable is way too Windowsesque for
> my taste.  But, it's back up and running for the users -- and that's
> what really matters.

In my experience, this is a function of the Berkeley Database that
is used.  I'm not sure precisely what is going on, but essentially,
there is no server process with Berkeley Database.  So if you die
while holding the locks, the locks stay held.  (Having no server
process is both the greatest feature, and biggest source of
shortcomings of BDB in my experience).

At this point, you have two options (at least this is what the
Subversion people talk about, they use BDB as the backend storage
for a SCM and end up having similar problems to RPM):

1.  Reboot to release the locks.

2.  Ensure no processes are actually accessing the database.  Run
the appropriate BDB function to release the locks.


This is what I've seen for option number 2 in RPM-parlance:

killall -9 rpm rpmb rpmd rpme rpmi rpmk rpmq rpmt rpmu rpmv; 
rm -rf /var/lib/rpm*.* /var/lib/rpm/__db* /var/lib/rpmrebuilddb*.*; 
rpm -v --rebuilddb &


As seen here:
http://beau.org/pipermail/whitebox-users/2004-May/001698.html

I believe this is the same set of commands that Vincent recommended
to me in another e-mail I can't seem to lay my hands on.  I'm not
sure I understand all of what it's doing, so if it breaks, you get
to keep the pieces.

	Thanks,
		Kirby

PS:  Vincent if you still read this list, thanks again.  It's saved
me plenty of trouble on several of occasions.