Unidentifiable Segmentation Fault

* 3500 10mb clients
* Dual PIII 500 Xeons w/ 1GB RAM and 2 NICs
* 7 cache dirs of 5GB a piece
* Linux (Slack 7) 2.2.14 (and 2.2.13)
* Squid 2.3STABLE1 with max_mem of 500MB
* compiled with --disable-ident-lookups --enable-heap-replacement*
        --enable-snmp --enable-dlmalloc*
(options followed by * have been included and excluded in different cfgs)

Squid operates just fine with 250 http requests per second. It will run
fine for days, but when i turn up the volume to say 1000 http requests/s
squid will log "FATAL: Received Segment Violation...dying" and instantly
shutdown and restart itself every couple of hours (or minutes depending
on the load). I have increased the amount of FDs from 1024 to 4096
and have also increased my ip_local_port_range to about 20k ports both
have had no effect. The last time it had died, there were only 400k
objects in the cache so it shouldnt be that... tried compiling with
different optimization args (-march=i686 or -O3, both, and niether),
verified that the rest of the system is running normal... everything
checks out. I just cant seem to figure this out and I havent seen any
answers to this(similar) question in the archives.

Hoping someone has some insight,
jk
Received on Wed Feb 23 2000 - 17:09:16 MST