cycbuffs and expire

Jeffrey M. Vinocur jeff at litech.org
Tue Sep 25 03:13:42 UTC 2001


On Tue, 25 Sep 2001, Alan Shackelford wrote:

> Then I am overlooking something obvious. We agreed that the cycbuff will
> only fill to capacity then articles start to flow out onto the floor. What,
> then needs to be expired? Is it just the headers? The articles are already
> gone. Since this is a database of sorts I suppose the headers may be the
> problem, and need to be expired to preserve the indexing (not the proper
> terminology, but I hope you get my meaning).

Oh.  Sure.  The overview entries for expired articles from cycbuffs will
be removed at the next (nightly) expireover run.


> We have just never had the article retention that we had hoped for,
> and I always blamed it on the cycbuffs filling up and articles were
> pushed out.

Sure...you can only keep as many articles as you have space for.


> Now that we are building big I had hoped to use the same "feature" to
> control article retention by carefully allocating space for the
> various hierarchies and setting expire for some absurdly high value
> and forgetting about it.

The expire settings for CNFS are ignored (except for the /remember/ line)
entirely.  Articles expire when they get overwritten in the cycbuff.

If you've got binaries, you definitely want to put them in a separate
cycbuff to keep them from pushing everything else out.

Some people use CNFS for binaries and a different storage method for their
other groups.


-- 
Jeffrey M. Vinocur
jeff at litech.org



More information about the inn-workers mailing list