Hiya,
We're running 2.1PRE1 on a FreeBSD 3.0 box. Every night when we see a peak
in usage, the following messages appear in the logs a couple of times
1998/10/27 21:38:07| comm_open: socket failure: (55) No buffer space available
1998/10/27 21:38:07| Reserved FD adjusted from 100 to 3083 due to failures
1998/10/27 21:38:07| WARNING! Your cache is running out of filedescriptors
leaving only some 400 usable filedescriptors for squid after a couple of
these messages.
We recompiled the kernel to have more mbufs available; netstat -m now
says:
2918/5472 mbufs in use:
573 mbufs allocated to data
2345 mbufs allocated to packet headers
535/3060/4096 mbuf clusters in use (current/peak/max)
6804 Kbytes allocated to network (21% in use)
824 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines
can anyone point me to what I'm not seeing here ? Which requests for
memory are being denied ? Should we change more kernelparams and if so
which ?
Cheers,
Frans
-- Frans ter Borg * * EuroNet Internet BV Network Operations * * Herengracht 208 - 214 * 1016 BS Amsterdam E-mail: frans@euro.net * Tel: +31 20 535 55 55Received on Tue Oct 27 1998 - 15:21:42 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:49 MST