> I'm running Squid-2.4.STABLE6 (in accelerator mode) on FreeBSD 4.4-RELEASE
> (dual P3 450Mhz)
>
> After 5 minutes of running, I'm getting this error in my cache.log:
>
> comm_open: socket failure: (55) No buffer space available.
>
> 2016 Kbytes allocated to network (65% of mb_map in use)
I had the same problem on FreeBSD 4.4 and 4.2,
when mb_map percentage wah high.
I wondered, because the peak value was lower that max!
I increased mbufs, and now I have
# netstat -m
101/288/16384 mbufs in use (current/peak/max):
101 mbufs allocated to data
99/274/4096 mbuf clusters in use (current/peak/max)
620 Kbytes allocated to network (5% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines
Add to /boot/loader.conf
kern.ipc.nmbclusters="4096" # Set the number of mbuf clusters
kern.ipc.nmbufs="16384" # Set the number of mbufs = 4 * nmbclusters
and reboot.
Or recompile the kernel and reboot.
Received on Tue Apr 16 2002 - 05:13:13 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:07:34 MST