Hi there, Could this help ? cache_peer <original_server_ip> parent <http_port> 0 no-query originserver name=<any_name> max-conn=4000 Kien Le. On Mon, May 16, 2011 at 1:34 PM, Or Gerson <OrG@xxxxxxxxxxxxxx> wrote: > Thanks for the help, > > Just some clarification to understand the logic of squid: " Squid uses as many connections to each peer as there are > maximum parallel requests needing to go to that peer" > > how is it that squid has a lot less connections to each peer than the number of connections he opens to each peers? > I mean he has 3 peers to divide the requests to. So I would have thought that the number of connections to each peer will > Be roughly the number of connections to squid divided by 3. > > How is the maximum parallel requests needing to go to that peer is much higher than the number of connections to squid? > > > > Or Gerson > IT Manager > > Mobile: 972-54-555-0656 Office: 972-3-769-8513 > E-mail: org@xxxxxxxxxxxxxx Website: http://www.websplanet.com > > > > > WebsPlanet, the leading provider of platforms for the mass production of websites > > > > -----Original Message----- > From: Amos Jeffries [mailto:squid3@xxxxxxxxxxxxx] > Sent: Monday, May 16, 2011 3:29 AM > To: squid-users@xxxxxxxxxxxxxxx > Subject: Re: squid too many connections to cache_peers > > On Sun, 15 May 2011 15:05:52 +0000, Or Gerson wrote: >> Hello everyone, >> >> I have squid 3.0.STABLE19-1 installed. >> And 3 apache cache_peers. >> >> I find that squid is opening a lot of connections to the web servers >> a lot more than the connections he is receiving from clients, also >> squid doesn't close them and they are in TIME_WAIT status. >> For example: in squid I see 324 TIME_WAIT connections >> On each server I see around 800 TIME_WAIT connections > > TIME_WAIT is a closing connection. > ESTABLISHED is an open connection. > > This explains: > http://www.developerweb.net/forum/showthread.php?t=2941 > > >> >> The cache_peers are configured with round-robin, but why squid >> doesn't closes the connections? >> >> Can't squid use one connection for each peer and send and receive all >> requests through that connection (although that probably will cause >> performance problems in the apache side because he won't uses all >> apache server processes)? > > One connection to each is indeed a performance bottleneck. Pipelining > is also not safe for HTTP/1.0 software (ie Squid 3.0 and older) without > great risk of breakage (just ask Debian people about apt-get corruption > problems). Squid uses as many connections to each peer as there are > maximum parallel requests needing to go to that peer. Enable > server_persistent_connections to servers and Squid will re-use > connections as much as possible given the HTTP support version. > > An upgrade to 3.1 series Squid will get you HTTP/1.1 connections to > servers and peers. This gives a lot better re-use of the connections > than HTTP/1.0 connection support in 3.0 series can. Provided Apache > plays along and uses HTTP/1.1 features to prevent early TCP link > closure. > > Amos > > This message, together with its attachments, contains information from > WebsPlanet Ltd., which is privileged and confidential. > If you are not the intended recipient or you have received this message in > error, then please notify us immediately by e-mail to info@xxxxxxxxxxxxxx, > and delete all copies of this message and its attachments. >
