[Yum] Yum 2.2.2 and FUTEX_WAIT

[Paul Nasrat]

On Wed, 2005-07-27 at 16:56 -0400, seth vidal wrote:
> On Wed, 2005-07-27 at 11:42 -0400, Brian Long wrote:
> > Hello,
> > 
> > We run yum 2.2.2 in %post of kickstart to install various yum groups.
> > We do it in a for loop like so:
> > 
> > for group in Insight-Manager ECS-Workstation ClearCase6 CrossOver-Office
> > Eclipse Enlightenment FVWM2 Oracle-Client Valgrind WindowMaker XFCE
> > Xine; do
> >     COMMAND="/usr/bin/yum -y groupinstall $group"
> >     echo "YUM COMMAND: $COMMAND" >> /var/log/postinstall
> >     $COMMAND </dev/null >>/var/log/postinstall 2>&1
> >     if [ "$?" != 0 ]; then
> >         echo "YUM FAILURE: $COMMAND" >>/var/log/postinstall
> >     fi
> >     /usr/bin/yum clean all
> > done
> > 
> > We have to do this because /var would fill up considering the size of
> > the groups (yum clean is needed after each group).  I'm watching an
> > install where the group FVWM2 installed fine, but Oracle-Client is hung.
> > When I switch to Alt-F2, strace on the yum process shows:
> > futex(0x8ee4bd8, FUTEX_WAIT, 1, NULL <unfinished> ...>
> > 
> > This yum process has been just sitting there for the last 25 minutes
> > with zero output on the screen.  It appears its waiting on a locked RPM
> > database or something.  There are zero RPM or yum commands running
> > besides this yum command.
> > 
> > What is going on?  Is there a timeout we can specify in yum.conf so yum
> > will fail after say 10 minutes of being unable to lock whatever it's
> > trying to lock?
> > 
> 
> cc'ing paul nasrat so he can explain a little better what the story is
> about FUTEX_WAIT.

What rpm version is this with?  Does any of the rpms within
Oracle-Client group do anything with rpm in it's scriptlets?

Rather than killing yum can you chroot and report

cd /var/lib/rpm
/usr/lib/rpm/rpmdb_stat -CA

Paul