Tip for those having ovdb/buffindexed problems.
mrcpu at internetcds.com
Tue Jan 30 16:08:22 UTC 2001
While I experiment with various versions of INN and ovdb and
do my darndest to break it, the rebuilding of ov is a major pain.
What I do now is have a duplicate feed of overview data go
to $spool/outgoing, and rotate it out every hour or so into a
backup dir. Then if I have to rebuild ov, I can
just pipe the last x hours of ov data that I want to recover
back to overchan, and I'm done.
WHich is a heck of a lot faster than makehistory.
Of course, it doesn't recover everything, but if you start
from oldest to newest, and you guess pretty close to your
CNFS buffer retention, you can end up only missing a few hours.
So far it's been working fine that I can tell, and
it only takes a few hours to recover, rather than a day or so.
Since for both buffindex and ovdb you can have multiple overchan's
it's not an issue, since the next expireover will clean up dupe's.
It may also help for tradindexed, depending on your spool mechanism
in reducing the wait time to bring the server back on line.
More information about the inn-workers