Hi Ingo,
Unfortunately I am stuck using FreeBSD 4.11 for this, so the em
driver options don't include rx_processing_limit. They do have rx and tx
_int_delay which I have played with but hasn't made a different.
As I understand polling might help if I had a load problem but my processor
is idle and it only happens after a day or more?
A netstat -na|wc -l gives over 7000 connections, but that shouldn't be a
problem? I have plenty mbufs and mbuf clusters available ..
Thanks
Dave
-----Original Message-----
From: Ingo Flaschberger [mailto:[EMAIL PROTECTED]
Sent: Tuesday, February 26, 2008 8:45 PM
To: Dave Raven
Cc: freebsd-net@freebsd.org
Subject: RE: Hping/Ping
Dear Dave,
> I have set my icmplim to 1024. When it was 200 I would sometimes get a
> warning about an OPEN port RST - which was also strange to me (that it was
> open not closed). I suspected that it was because 5% of my packets where
> being dropped though.
>
> The server is under reasonable load (networking wise) around 80-90mbits,
but
> the processor is 75% idle..
>
> The other strange this is that a reboot solves this, and it takes around
1-2
> days to happen again..
you can try to enable polling:
kern.polling.reg_frac=20
kern.polling.user_frac=10
kern.polling.burst_max=1000
and perhaps tune rx_processing limit:
dev.em.0.rx_processing_limit: 100
you can also look at all open sessions when the server is in trouble:
netstat -n
Kind regards,
Ingo Flaschberger
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
