squid stalls for 5-10 minutes

Since yesterday, I've been running a script that does a cache_object
request to the server every 5 minutes. Three times this morning, the
server stopped accepting connections (both regular HTTP and my
cache_object connection) for between 5 and 10 minutes, as my script would
miss two consecutive polls.

There is nothing unusual in the access or log files, other than a gap in
time where there are no entries.

My configuration is a single server (no neighbors or parents), clean_rate
is off by default, and most cache parameters at their default values,
except I have setup 16 cache_dir's just to split things up. The server is
a Sun SPARC 5 running running SunOS 4.1.3_U1 with 128 MB of RAM and a 200
MB cache_swap. We get about 200,000 connections to the proxy server a
day.

I started my script about 5:15 yesterday, and the stalling only happened
at 9:40, 10:40 and 11:45 this morning. I sent a SIGSEGV signal to the
server at about 11:50, hoping for a core dump, but didn't get one. Since
restarting the server, the stalling has not returned.

Over the last few months, I have noticed that when running either cached
1.4 or squid 1.0.x, that the server would occassionally lock up for longer
than the 10 minutes I'm noticing now. For cached, we had to restart the
server several times a day, while for squid, it is a couple of times a
week.

Anyone seen this or know how to fix this?
------------------------------------------------------------------------
Edward Moy
Xerox Palo Alto Research Center
3333 Coyote Hill Rd.
Palo Alto, CA 94304

Email: moy@parc.xerox.com
WWW: http://www.parc.xerox.com/moy/
PGP key fingerprint: AA A1 12 00 9B 13 07 45 19 61 26 A1 AF AF 99 F3
Received on Wed Jul 31 1996 - 14:48:26 MDT