[SQU] Thrashing squid

I've been running Squid 2.3 under FreeBSD 4.1 since the end of August.
The machine I'm running it on is a PII300 w/96MB RAM and a single 8 GB
IDE disk. I started it out with a 1GB cache and after getting some SNMP
monitoring on the cache, I gradually raised it up to 2.5GB of cache.
>From swap usage, it appeared that I was at the edge of cache size vs
physical RAM, but overall performance was "good enough" until next
fiscal year.

Due to a UPS upgrade, I had to shutdown the cache this weekend -- the
cache system was halted and powered off gracefully and then powered back
up when the UPS came back online.

After being power cycled, the cache seemed really, really sluggish --
pages pulled from sites like Yahoo took 2-3 seconds to load. Upon
further inspection, the cache machine is thrashing its disk like crazy.
I backed the cache size down 512MB at a time to 1GB with no effect
(rebooting between squid.conf changes, and letting it thrash for a
couple of hours to see if it was a temporary thing). Even after halting
squid, rm'ing the cache itself and rebuiling it with squid -z and
restarting it was still thrashing like crazy.

What am I missing here? I'm kind of lost as to why squid would suddenly
start using tons of memory -- in excess of 90MB according to top -- on
an empty cache dir and with the only clients users who left their PCs on
over the weekend with browsers on web pages with REFRESH tags..

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html