I am using Squid 2.4Stable3 on FreeBSD 4.5. I have been getting this error over and
over in cache.log:
comm_open: socket failure: (55) No buffer space available
when doing 'netstats -m' I noticed that I had peaked out at my max mbuf clusters, so I
recompiled the kernel with more mbuf clusters. I also upped kern.maxfiles and
kern.maxfilesperproc. Finally, I recompiled squid. However, I am still getting the error.
Do I need to shrink the size of the cache on disk so that squid uses less RAM perhaps?
Perhaps squid doesn't see the new buffer space available? I am planning to get more
RAM for this system, but here is its setup:
P3-600, 128MB PC133. I have a disk cache of 5GB, the disk is IDE and is using
UltraATA-33.
'netstat -m' reports the following:
--- 130/976/16384 mbufs in use (current/peak/max): 130 mbufs allocated to data 128/956/4096 mbuf clusters in use (current/peak/max) 2156 Kbytes allocated to network (17% of mb_map in use) 0 requests for memory denied 0 requests for memory delayed 0 calls to protocol drain routines --- Thanks, LucasReceived on Thu May 16 2002 - 12:21:19 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:08:08 MST