I have deployed a field test of a squid 3HEAD box running TPROXY in a WCCP configuration. Under more client load (that could not be simulated in my lab) I am noticing that users are getting "(99) Cannot assign requested address" errors from the squid box. The odd thing is that some web sites cause the error to happen almost all the time when the user gets to a certain point through the website. Additonally the same sites work fine in Safari on MacOS 10, but not in IE6 on WindowsXP SP2 (and SP3). An observation from one enduser noted that if they hit the refresh button in the browser enough, the error goes away. Some sites that are giving problems are: www.surveymonkey.com shl-isp.com bestkits.com metraonline.com I am using the following software versions: CentOS 5.2 (x86_64) squid/3.HEAD-20080721, with tproxy-onresetfd.patch from Amos Jeffries (back on 7/26/08) iptables-1.4.0, with tproxy-iptables-20080204-1915.patch kernel 2.6.25-11, with tproxy-kernel-2.6.25-20080519-165031-1211208631 I am still getting the following entries in my cache.log, but not super often since setting] "echo 1 > /proc/sys/net/ipv4/ip_nonlocal_bind" in /proc. 008/09/09 14:55:22| IPInterception.cc(137) NetfilterInterception: NF getsockopt(SO_ORIGINAL_DST) failed: (11) Resource temporarily unavailable 2008/09/09 14:55:22| IPInterception.cc(171) NetfilterTransparent: NF getsockopt(IP_TRANSPARENT) failed: (92) Protocol not available My squid.conf uses this for the http_port setting: http_port 3128 tproxy intercept disable-pmtu-discovery=always My question is, what can I do to troubleshoot this, or give members of this list more information to help isolate the problem? Does anyone have suggestions? I can post any information as requestion regarding the specific configuration, etc.
