mpd5/Netgraph issues after upgrading to 7.4

[Przemyslaw Frasunek]

W dniu 08.04.2011 22:13, Przemyslaw Frasunek pisze:
> I'm still looking for a help in investigating this issue. The problem appears on
> two 7.4 boxes, while 7.3 are working OK. Ktrace shows, that indeed some of
> sendto() calls on netgraph control socket are failing with EPERM:
[...]

Eventually I found that this issue is related to mbuf exhaustion. In
periods when sendto() fails with EPERM, the "requests for mbufs denied"
counter is increasing and "ngctl list" also fails.

Increasing kern.ipc.nmbclusters (which currently equals to 262140)
doesn't seem to fix the problem.

[root at lsm-gw /var/log]# ngctl list
ngctl: send msg: Cannot allocate memory

[root at lsm-gw /var/log]# netstat -m
2068/12317/14385 mbufs in use (current/cache/total)
2067/2325/4392/262140 mbuf clusters in use (current/cache/total/max)
6/694 mbuf+clusters out of packet secondary zone in use (current/cache)
0/0/0/12800 4k (page size) jumbo clusters in use (current/cache/total/max)
0/0/0/6400 9k jumbo clusters in use (current/cache/total/max)
0/0/0/3200 16k jumbo clusters in use (current/cache/total/max)
4652K/7729K/12382K bytes allocated to network (current/cache/total)
3610/43753/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)
0/0/0 requests for jumbo clusters denied (4k/9k/16k)
0/34/6656 sfbufs in use (current/peak/max)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
0 calls to protocol drain routines