Hi Henrik, I applied the workaround you suggested and did a quick test. It seems working. I would like to know what's the impact of this change. >Hmm.. if you see a long delay between Squid sent the request and >ERR_ZERO_SIZE then something is wrong at the web server, eating the >request without responding to it. Yes, I do see 1 mins delay between Squid sent the request and ERR_ZERO_SIZE. (For cache.log:) 2006/08/17 16:12:03| httpSendRequest: FD 19: GET /conformance/vxml-srgs-voicegenie/278/record5-pass.vxml HTTP/1.0^M Host: osiris.voicegenie.com:8080^M Cookie: JSESSIONID=F1E33EC3C461E6AA42AFF57784A47B7B^M Accept: application/grammar-xml, application/octet-stream, application/x-abnf, application/x-javascript, application/x-jsgf, application/x-ms-xml, application/x-swi-grammar-compiled, application/x-voicegenie-nuance, application/x-voicegenie-watson, application/srgs+xml, application/ccxml+xml, audio/L16, audio/L8, audio/adpcm, audio/basic, audio/vox, audio/wav, audio/x-adpcm, audio/x-adpcm8, audio/x-alaw-basic, audio/x-basic, audio/x-g726, audio/x-g726-16, audio/x-g726-24, audio/x-g726-40, audio/x-vox, audio/x-wav, text/html, text/plain, text/vxml, text/x-vxml, text/xml^M X-Session-Id: 0002004E-10479A7A-0001^M Pragma: no-cache^M Referer: http://osiris.voicegenie.com:8080/conformance/vxml-srgs-voicegenie/278/cgi-bin/record5.ircgi^M User-Agent: VoiceGenie NXP/255.255.255^M Via: 1.0 morgan.voicegenie.com:3128 (squid/2.5.STABLE5)^M X-Forwarded-For: 127.0.0.1^M Cache-Control: max-age=259200^M Connection: keep-alive^M ... ... 2006/08/17 16:13:02| fwdFail: ERR_ZERO_SIZE_OBJECT "Service Unavailable" http://osiris.voicegenie.com:8080/conformance/vxml-srgs-voicegenie/278/record5-pass.vxml Do you have any clue what is wrong here? Thanks, Katie ________________________________ From: Henrik Nordstrom [mailto:henrik@xxxxxxxxxxxxxxxxxxx] Sent: Thu 9/7/2006 1:22 PM To: Xin Wang Cc: squid-users@xxxxxxxxxxxxxxx Subject: RE: Why ERR_ZERO_SIZE_OBJECT? tor 2006-09-07 klockan 11:16 -0700 skrev Xin Wang: > The ERR_ZERO_SIZE was not sent to the requesting client. However, > since the squid takes 1 minute to respons (due to ERR_ZERO_SIZE_OBJECT > and retry), the requesting client timed out. The requesting client is > actually our customer's application, and 1 minute is not acceptable to > the application. Didn't see any delay in the trace data, but maybe I overlooked something.. The retry was within a few ms after the ERR_ZERO_SIZE. Hmm.. if you see a long delay between Squid sent the request and ERR_ZERO_SIZE then something is wrong at the web server, eating the request without responding to it. You can most likely work around the issue by "server_persistent_connections off". Will cost a little in added latency, but if the serer is close to the proxy then it's marginal. Regards Henrik
