Scenario: I configure a squid cache server as a transparent reverse proxy based on the instruction in http://squid.visolve.com/white_papers/reverseproxy.htm When users type a URL on their browsers, the local dns server maps the host address to the proxy server and the request is sent to the proxy server, then the proxy server retrieves the content and sends back to the browsers. But when they type a valid url and retrieve the content successful, and then switch to a url pointing to a bogus web server, the squid server generate core dumps constantly. I notice when the core dump happens, the debug option in the conf is debug_options 33,9 If I change it to debug_options 21,9 I don't see squid crash any more, does anyone know why? Environment: Ultrasparc running Solaris 8 (Generic_108528-18 sun4u sparc SUNW,UltraAX-i2) and squid version is 2.5 release 4 Gdb stack: #0 0xff01f7cc in _libc_kill () from /usr/lib/libc.so.1 (gdb) where #0 0xff01f7cc in _libc_kill () from /usr/lib/libc.so.1 #1 0xfefb58f4 in abort () from /usr/lib/libc.so.1 #2 0x6adf4 in sigusr2_handle (sig=11) at tools.c:303 (gdb) Squid Conf file: Here is the diff file between the modified squid conf file and original conf file. < # http_port 3128 > http_port 80 > http_port 8080 < # https_port > https_port 443 cert=/usr/local/squid/ssl_gw.pem < # cache_peer cache.domain.example parent 3128 3130 default > cache_peer 192.168.0.2 parent 3128 3130 < # httpd_accel_host virtual > httpd_accel_host virtual < # httpd_accel_port > httpd_accel_port 0 > httpd_accel_host virtual < # httpd_accel_single_host off > httpd_accel_single_host off < # httpd_accel_with_proxy off > httpd_accel_with_proxy on < #always_direct allow acceleratedHost < #http_access allow acceleratedHost acceleratedPort < # cache_mem 8 MB > cache_mem 50 MB < # httpd_accel_uses_host_header off > httpd_accel_uses_host_header on < # cache_replacement_policy lru > cache_replacement_policy lru < # memory_replacement_policy lru > memory_replacement_policy lru == Sean Brennan
