Examples on how others use expire/expirerm/fastrm ?

Russ Allbery rra at stanford.edu
Wed Apr 12 02:44:02 UTC 2000


BARRY BOUWSMA IS AN INFLUX OF CONTUMELIOUS FINKS writes:

> We're running a full feed on 14 day remember time, with a history text
> file just topping 1,1GB now, and about 130+90MB for the database files.
> Non-normal k0deZ so these filesizes are much smaller than for Real
> people.

> In spite of this, I routinely see that many of the sites I feed reject
> articles we offer as being `Too Old'.  And worse, while I was setting up
> several new test machines, keeping the /remember/ of 14 days and copying
> the old text file to the new machines, I still saw quite a few log
> entries that *we* were rejecting some articles seen for the first time
> as being too old, since the default cutoff had been only 7 days by
> default, and I didn't catch this until later.

> Quite some time ago, I did some counts of messages being received that
> were not yet in history, with different cutoff times.  I'd say you will
> see several tens of thousands of new `too old' messages with a seven-
> day cutoff -- even with ten days or more, I saw more rejected messages
> than I cared to see when I was making these tests.

Well, I'm currently running with a 7-day cutoff, and I'm not seeing a huge
number:

Server                   Total  Group Dist Duplic Unapp TooOld Site Line Other
TOTAL: 35               772639 573715    0  56354   395    213    0    0 141962

Of course, I don't try to accept every hierarchy, and more obscure
hierarchies may have more of a problem with this.  213 too-old articles
doesn't seem like much to me.

But in the absence of anyone else strongly agreeing with changing the
defaults, I'll leave them as is for now.  2.4 should hopefully see
sufficient architectural changes to the history mechanism to make this
less of a problem.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>



More information about the inn-workers mailing list