Network buffer problem

[Sascha Holzleiter]

Hi,

i've bit of an understanding issue with the netstat output which seems
to condradict itself a bit.
One of our production machines has a moderate network load (24mbit on a
100mbit device) and quite frequently shows a packet loss rate around 2%
on peak times.

A netstat -s -p ip output looks like this:

        1459405058 packets sent from this host
        0 packets sent with fabricated ip header
        411987593 output packets dropped due to no bufs, etc.

So there are a lot of packets dropped due to buffers.
But the mbuffer statistic with netstat -m shows no sign of this:

3470/2095/5565 mbufs in use (current/cache/total)
1536/1078/2614/25600 mbuf clusters in use (current/cache/total/max)
1536/145 mbuf+clusters out of packet secondary zone in use
(current/cache)
0/0/0/0 4k (page size) jumbo clusters in use (current/cache/total/max)
0/0/0/0 9k jumbo clusters in use (current/cache/total/max)
0/0/0/0 16k jumbo clusters in use (current/cache/total/max)
3939K/2679K/6619K bytes allocated to network (current/cache/total)
0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)
0/0/0 requests for jumbo clusters denied (4k/9k/16k)
0/7/6656 sfbufs in use (current/peak/max)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
5646 calls to protocol drain routines

There are no mbufs denied ever, the mbuf max is far above the total
amount used so i'm a bit at a loss what's the problem here but it seems
to be some kind of buffer issue.

I already raised kern.ipc.maxsockbuf to 16777216 and the nmbclusters is
still at 25600 which seems fine according to the mbuf statistic above.

This is on a GENERIC FreeBSD 6.2 system with an em network card at
100mbit. 


Greetings,
  Sascha