Hi, On 05/08/08(21:49) you Dmitry Donskih <ppyhqbdyi-5gwnve4piqdw.ml@delegate.org> wrote in <_A3966@delegate-en.ML_> |I use DeleGate 9.2.1 as HTTP proxy on Fedora box. Most its settings |(timeouts etc) are default. |A client requests a big file (http GET, >700 Mbytes). DeleGate starts |retrieving it from server. Then this client (by HIS words) disconnects. |Somehow proxy does not terminate, and downloads entire file. | |(We pay for internet traffic, and this file is being disputed) | |Please help me to find out if this client really has disconnected at the |very beginning of download, or not. | |If so, then why did DeleGate continue downloading that file. Is this a |timeout/takeover issue, or what? The default value of "caching takeover" is just 6 seconds but it took longer than 3 minutes in your case. |05/07 17:46:26.66 [9974] 2345+8: HTTP/1.1 200 Content-{Type:text/html Encoding:[/chunked] Leng:0} Server:Apache/1.3.37 (Unix) mod_ssl/2.8.28 OpenSSL/0.9.7e-p1 FrontPage/5.0.2.2510 |05/07 17:46:32.13 [9974] 2345+8: detach respBuff: huge data |05/07 17:46:32.27 [9974] 2345+8: discard cache: leng=1049280/0 /var/spool/delegate-nobody/cache/http/krugmebeli.ru/index.php?catalog=1032 >>> 3 minutes <<< |05/07 18:49:42.23 [9974] 2345+8: ## premature client close: flush_body (cant_getpeername) |05/07 18:49:42.23 [9974] 2345+8: ClientEOF: flush_body |05/07 18:49:42.23 [9974] 2345+8: HTTP transmitted: 275head+000000000/fbody=>1049280txt+0bin->734250566/0, 807026i/47843o/0f/3795.6 In the above silent period, maybe DeleGate/9.2.1 is receiving the data into a temporary file to do gunzip/gzip the data, without relaying it to the client. DeleGate tries detecting the disconnection from the client dureing the buffering, but it might fail in some reason. I recommend you to use recent version of DeleGate in which the above buffering for gzip is replaced with the pipelined gzip/gzip threads. Cheers, Yutaka -- 9 9 Yutaka Sato <pfqcabdyi-5gwnve4piqdw.ml@delegate.org> http://delegate.org/y.sato/ ( ~ ) National Institute of Advanced Industrial Science and Technology _< >_ 1-1-4 Umezono, Tsukuba, Ibaraki, 305-8568 Japan Do the more with the less -- B. Fuller