[Mageia-dev] Fatal errors with rpm

[Robert Fox]

On Wed, 2012-03-14 at 18:19 +0100, Wolfgang Bornath wrote:
> 2012/3/14 Wolfgang Bornath <molch.b at googlemail.com>:
> > 2012/3/14 Wolfgang Bornath <molch.b at googlemail.com>:
> >> 2012/3/14 Wolfgang Bornath <molch.b at googlemail.com>:
> >>> 2012/3/14 Wolfgang Bornath <molch.b at googlemail.com>:
> >>>> 2012/3/14 Robert Fox <list at foxconsult.net>:
> >>>>> On Wed, 2012-03-14 at 16:10 +0100, Olav Vitters wrote:
> >>>>>> On Wed, Mar 14, 2012 at 04:01:03PM +0100, Robert Fox wrote:
> >>>>>> > Since the last few updates, I am experiencing erratic RPM db errors (on
> >>>>>> > more than one machine):
> >>>>>>
> >>>>>> This should already be fixed. Maybe your mirror lags behind a bit.
> >>>>>>
> >>>>>
> >>>>> I have the latest updates and it just happened - I will rebuild the db
> >>>>> and report back if it happens again.
> >>>>
> >>>> Just did updates (including rpm-4.9.1.2-22.mga2) on 2 cauldron
> >>>> installations. Last mirror sync with tier1 was 1:20 hours ago.
> >>>> No problems with rpm db
> >>>
> >>> Now I did some more package management:
> >>>  - I opened a konsole window, did another check with 'urpmi
> >>> --auto-update' which turned up no updates. After that rpm database was
> >>> unlocked. Good.
> >>>  - I opened rpmdrake from the konsole and marked all KDE handbooks for
> >>> uninstall. I clicked on OK and watched the output in the konsole
> >>> window.
> >>>  - after the second handbook urpmi crashed with a segmentation fault
> >>> and rpmdrake showed a message box with the same error (segfault).
> >>>
> >>> Then I rebooted the system and starting rpmdrake on the konsole returned:
> >>> * getting lock on urpmi
> >>> * [reading all synthesis files, ok]
> >>> * Error: rpmdb: Thread/process 2794/1397..... failed: Thread died in
> >>> Berkeley DB library
> >>> * Error: db4 error from dbenv->failchk: DB_RUNRECOVERY: Fatal error,
> >>> run database recovery
> >>> * Error: Unable to open packages index, using db4
> >>> * Error: Unable to open package database in /var/lib/rpm
> >>> * unlocking urpmi database
> >>>
> >>> Mind that this happened with latest updates as written in my previous mail.
> >>
> >> Adding info: erasing and rebuilding the db does not work. After
> >> removing the _db* I run # rpm --rebuilddb which returns
> >> :
> >> * Error: rpmdb: Thread/process 2794/1397..... failed: Thread died in
> >> Berkeley DB library
> >> * Error: db4 error from dbenv->failchk: DB_RUNRECOVERY: Fatal error,
> >> run database recovery
> >> * Error: Unable to open packages index, using db4
> >>
> >> Result: system is unusable now.
> >
> > Solution: I had to manually erase the lock file /var/lib/rpm/.RPMLOCK
> > After that rebuilding the database worked and urpmi is usable again.
> 
> Ok, here we go again:
> I did the same as described earlier:
> Updates, then opening rpmdrake and marking all handbook packages for
> uninstall. After uninstalling one or two packages rpmdrake segfaults.
> Opening it again shows the known errors about the rpm database.
> 
> Unfortunately rpmdrake gives no other error message than "Segmentation
> fault" - that's it. So I don't know whether the error messages about
> the rpm database are just a result of the segfault or the cause of the
> segfault.
> 
> --
> wobo

Hello Wolfgang -

Try the rebuild again:
rm -f /var/lib/rpm/__db*
rpm --rebuilddb 

and then simple do:
urpmi --auto-update -v

That worked for me.  When I tried rpmdrake, I also got the corruption
again.

Cheers,
R.Fox