At 05:52 PM 12/27/00 +0100, you wrote:
>Hello Chris,
>
> > They are dropped if the multi part buffer runs out of space before all the
> > parts
> > of that particular message have arrived.
>
>Is there a way to tell how much space of the part buffer is in use? I 
>don't really understand the status output....

Yes it's the first figure on the line from memory, however the figure 
doesn't mean much as it's a circular buffer it just writes in a circle 
regardless of how much is
used (as it's more efficient)  so the rotate time is the figure to check.

> >So there are in theory only 4
> > explanations
> >          1) The incoming feed is not complete
> >          2) The system is not fast enough when the multi part buffer is
> > turned on
> >          3) The multi part buffer is too small
> >          4) Something else :-)
> >
> > It sounds to me like '3' is the most likey candidate at the moment, so 
> if you
> > have the space increase the multi part buffer size.
>
>Ok I got that. I currently have 4.5 GB allocated for the multipart buffer. 
>You once told me that 2 GB would be way enough. I am going to increase it 
>to some 10 Gigs or so and see whether that helps.

I may have been wrong :-)


> > Second you have a 4% rejection figure, examine dnews.in or parterr.log to
> > see why those parts are being rejected it may be you need to modify a 
> spam rule
> > or something.
>
>I think I know why the 4% rejection happened and it may be a bug in DNews: 
>I had several high volume groups deleted, but not yet purged. Whenever an 
>article was posted that was also crossposted to one of the deleted groups, 
>then DNews rejected the article. At least that were the only rejection 
>messages I could see in parterr.log and there were plenty of them (roughly 
>15 MB of parterr.log every day!). I undeleted all those newsgroups and now 
>I have 0% rejection.
>
>So if an article was posted to two groups and one of the groups was marked 
>as deleted - DNews rejected it. Bug???

No, it's simply counting posts to deleted groups as rejects as it collects 
the parts before it realizes that they are for a group you don't want.  It 
would
be more efficient if it did this check the other way around, we'll look 
into that.

> > Lastly do you have thread_in true ? If not I'd add that.
>
>Isn't that turned on by default in the latest versions? My status output 
>says something about threading, so I suppose it is on:
>
>     Input threads running 8, output threads 0

Yes. It is turned on for some platforms in the latest version. And certainly if
that status line shows up.

                 Chrisp.

Reply via email to