i had this problem a month ago. after reading up code and
discussing with Rui Paulo,
i increased the number of mbuf clusters and it apparently
went away. i can't really
precise if it's really that correlated but try it and give
some feedback here again.
>The only uncommon thing about the machine in question is
that it has two
>network interfaces connected to the same ``wire''. We
run two IP nets
>(131.220.120/24 and 131.220.146/24) on the same ethernet
(until I find the
>time to partition the switches) and this was the way to
make DHCP work.
My machine had two interfaces too, one crossed-over to a
cisco router and another
one on a huge broadcast domain. both interfaces have alot
(10+) ip aliases and
they both are gateways for 1000+ other hosts. (i know,
horrible topology.)
the only reference to this kind of problem i found on the
web was some guy with
a sun sparcstation (or was it an ultra) and no one gave
feedback.
best regards,
Francisco Guerreiro
|