On Thu, 2009-05-21 at 21:06 -0400, Ross Walker wrote:
> On Thu, May 21, 2009 at 12:13 PM, JohnS <jses27@xxxxxxxxx> wrote:
> > On Wed, 2009-05-20 at 09:30 -0400, Ross Walker wrote:
> >> I wonder if yum-updatesd might cache repo data separate from yum
> >> cache, in which case some older incompatible cached data from the
> >> previous version may still be around causing yum-updatesd to bork.
> >>
> >> If so I think that cache will need to be manually deleted.
> >>
> >> I'll check my desktop system when I get to work to see if my theory is
> >> correct.
> >
> > The recent problem with having to yum clean all I have gone to the
> > extent of testing on a client machine. What i have done is run a cron
> > job for yum clean metadata whith a if then yum update. Evidently this is
> > not happing to everyone.
>
> You don't need a cron job for this issue, it's a one-time event as the
> repo metadata for the new yum isn't compatible with the old yum and if
> there was cached old metadata left over after the upgrade, that wasn't
> marked stale, then it would cause yum grief. Clearing out the metadata
> was the simple fix and only needed to happen once.
>
> I forgot to check to see if yum-updatesd kept a separate metadata
> cache too that might need manual clearing out.
>
> -Ross
---
from: /usr/sbin/yum-updatesd
def invalidate_cache(*args):
global updateInfoDone, updateInfo, helperProcess
if helperProcess is not None:
return
updateInfo = []
updateInfoDone = False
def setup_watcher():
"""Sets up gamin-based file watches on things we care about and
makes
it so they get checked every 15 seconds."""
# add some watches on directories.
mon = gamin.WatchMonitor()
mon.watch_directory("/var/lib/rpm", invalidate_cache)
mon.watch_directory("/var/cache/yum", invalidate_cache)
map(lambda x: os.path.isdir("/var/cache/yum/%s" %(x,)) and
mon.watch_directory("/var/cache/yum/%s" %(x,),
invalidate_cache),
os.listdir("/var/cache/yum"))
mon.handle_events()
fd = mon.get_fd()
gobject.io_add_watch(fd, gobject.IO_IN|gobject.IO_PRI,
lambda x, y: mon.handle_events())
------------------------------------------------------------------
Turns out that gamin is the one doing this metadata upkeep. Maybe Gamin
is not working right on some installs?
Ross, the reason I done the cron job way was because I have had these
problems before. From the code above it uses the same cache.
JohnStanley
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
