Stefan F?rster:
> * Wietse Venema <wie...@porcupine.org>:
> > I'm burning in some new code that I wrote over the past week to
> > periodically remove old entries from postscreen(8) and verify(8)
> > caches. This is not a sexy topic, but it helps to keep database
> > sizes in check, it speeds up database access, and it was time.
> >
> > If you're courageous you can try postfix-2.7-20091227-nonprod. If
>
> I had some problems integrating the Debian patches which support
> dynamic loading of map types (worse, I don't even understand why), so
> I went for a traditional compile and install.
You may want to give a heads-up to the maintainer as I am just
about to release this code as a regular experimental release.
Cache cleanup will definitely be part of the Postfix 2.7 stable
release, so they will have to deal with it in a month or so anyway.
> It's working as expected, I didn't see any adverse side effects so
> far. OTOH, the servers where I deployed the code have almost no real
> workload at this time between the years, possibly rendering my
> observations flawed.
>
> What kind of data could we provide to help you with this?
I think the only stats of interest are the cache cleanup statistics
which are logged twice a day by default. Mine have stabilized around
the following numbers:
Dec 29 04:20:17 spike postfix/postscreen[44900]: cache
/var/lib/postfix/ps_cache.db full cleanup: retained=134 dropped=19 entries
Dec 29 06:19:33 spike postfix/verify[46072]: cache /var/lib/postfix/verify.db
full cleanup: retained=1726 dropped=28 entries
I had some fun when I ran it on a verify cache file that had
accumulated about 300k entries. Cache cleanup took a couple minutes
while at the same time responding to routine lookup/update requests.
This is on an older machine.
Wietse
