expireover & tradspool

Katsuhiro Kondou Katsuhiro_Kondou at isc.org
Mon Jul 22 11:07:12 UTC 2002


In article <20020722094141.A16037 at fantomas.sk>,
	"Matus \"fantomas\" Uhlar" <uhlar at fantomas.sk> wrote;

} I think I should have no(less?) problem with disk i/o. I use expire on
} newsgroup names so this is also something very ineffective, it seems.
} 
} may I know, how does expire/expireover work in this cases? is there
} something that makes it use so much memory?

In your case, you seem to set groupbaseexpiry to true.  The
senario of expiry is(not generally, but this is for your
case);

- expireover
   - read all overview data for token, Message-ID and Xref
   - examine the history to see if the Message-ID of article
     exists(go to next overview data if not)   (a)
   - Xref is compared with configuration(expire.ctl) to see
     if article should be purged or not
   - append a pathname to a file(*) if so(the file is used by
     fastrm)
- sort file(*)
- fastrm
   - purge article with the file(*)
- expire
   - read each entry
   - examine the token to see if the article still exists (b)
   - drop token from the entry if not

At (a), dbz is read and you may need memory to keep, and also
checks the existence of all articles listed in history.  At
(b), it also checks the existence of all articles in history.
-- 
Katsuhiro Kondou


More information about the inn-workers mailing list