overview troubles with Solaris-8 and 2.3.2
dickon.hood at bbc.co.uk
Tue Jun 12 20:36:53 UTC 2001
On Tue, Jun 12, 2001 at 15:47:05 -0400, Mark G. Thomas wrote:
: Any suggestions about the below?
: I'm still having trouble with buffindexed ovmethod and 2.3.2 on my
: Solaris-8 box.
: After my original upgrade from inn-1.72 to 2.3.2, I had trouble
: with expireover core dumping. I then remade my overview buffers,
: re-created my history and overview data a second time (with
: makehistory and makedbz again), all seemed fine for a couple weeks, but then
: all of a sudden expireover went from taking 3-6 hours to taking almost an
: entire day, for the past couple days!
This is interesting.
A longish time ago (October or so last year, IIRC), we installed a fairly
major project on the website: the Central User Database. Part of that CUD
is based on Berkeley DB files (used via a Perl process or two). These
worked perfectly happily on Solaris 7, but oddly corrupted when moved to
sol. 8. We don't know why - the one person who might have a clue jumped
ship to work for a bunch of beancounters in central London...
Now I see people complaining here about overview database problems on Sol.
8 machines, which appear to be Berkeley DB related. I myself have been
suffering problems using Solaris 8 and inn 2.3.1 (again, odd ovdb errors:
mycw xovers have been losing articles for no very good reason that I can
see; request a missing article by article number and it's there - it's
also present via message ID). My workaround is to rebuild the database
every night (0400 is a reasonable time, I feel), but then I'm running a
system which isn't even Usenet...
Oh, and can anyone tell me if I should run expireover with inn throttled
or not? The manpage is unclear. My gut feeling is it should be, but I'm
not ATM - this is another potential cause of corruption...
BBC Internet Services,
(+44 1737 8)39754. Mail me. Don't phone if possible. Ta.
More information about the inn-workers