On May 2, 2002 09:15 am, Edgar Friendly wrote:
> Benjamin Coates <coates at windmail.net> writes:
> > The top of my OCM contents:
>
> <SNIP>
>
> > Notice the connection that's been running for 32 hours for one message,
> > and a few nearly as bad.
>
> Interesting.  Since we're pretty certain that the code to choose the
> most unused connection works, I want to conclude that for some reason
> your node hasn't found it necessary to drop any connections in the
> last 32 hours.
>
> > Also, while my node is getting 6-8 thousand requests per hour, it has
> > rejected all but one of them in the last 24 hours (at about 3AM), which I
> > would take to mean my node is overloaded despite doing nothing.
> >
> > --
> > Benjamin Coates
>
> Nodes seem to be doing a good job of returning QRej messages when
> they're overloaded, and completely rejecting connections when they are
> way overloaded (let's call it swamped), but it seems that after some
> time of running their ability to handle normal requests/inserts goes
> down the drain.  Does restarting your node restore its ability to
> handle new traffic?  And how many successful requests have gone
> through your node since it started?

After running a modified node here (routing backoff of 6 seconds instead
of 60, the accept/reject code tries to favor request we are likly to be able
to handle and I track the accept, Fail, and answer rates.  After 2.5 days 
uptime I see:

Rates: Accepting 29.48% Failing 0.76% Answering 0.58% of requests

with about 5500 requests per hour.   A 1.5 days ago I was accepting
50%, failing to reply to 4-6% and answering 3-4%...  Then my routing
table had 70 entries.  It now down to 12 with 11 backed off...

Ed Tomlinson







_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to