# What exactly does keepalive\_timeout work? **URL:** \ **Category:** NGINX\ **Created:** [June 17, 2016, 10:49am UTC](https://www.ruby-forum.com/t/what-exactly-does-keepalive-timeout-work/244801 "2016-06-17T10:49:38Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![Rafa\_F](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/958977/32.png) [@Rafa\_F](https://www.ruby-forum.com/u/Rafa_F)\ **Post date:** [June 17, 2016, 10:49am UTC](https://www.ruby-forum.com/t/what-exactly-does-keepalive-timeout-work/244801/1 "2016-06-17T10:49:38Z")
Hi, I’d like to understand how `keepalive_timeout` works. The description in the documentation isn’t very clear and I couldn’t get an answer to my satisfaction on #nginx IRC. People seem to have their own understanding of how `keepalive_timeout` works and they are very different. 1. For e.g. [https://github.com/roots/trellis/blob/master/roles/nginx/templates/nginx.conf.j2](https://github.com/roots/trellis/blob/master/roles/nginx/templates/nginx.conf.j2) describes the directive like this: better # for each individual client, particularly for SSL, but means that worker # connections are tied up longer. (Default: 65) So according to the statement `keepalive_timeout` is to be used to tell how long a server needs to keep the connection alive AFTER serving all the requests. And this commenter seems to agree: [http://serverfault.com/q/331762#comment665957\_331764](http://serverfault.com/q/331762#comment665957_331764) > [It’s] only a timeout during which server waits for another request > up to [the limit set by] `keepalive_requests`. So it is not important > how much time it takes for the full page to load […] BTW the max. > time connection is kept open is almost: > > ``` > keepalive_timeout * keepalive_requests > > ``` 1. Then there are people understand the function of `keepalive_timeout` like so: [HTTP Keep Alive and TCP keep alive - Stack Overflow](http://stackoverflow.com/a/10202959) > HTTP Keep-Alive is a feature of HTTP protocol. The web-server, > implementing Keep-Alive Feature, has to check the connection/socket > periodically (for incoming HTTP request) for the time span since it > sent the last HTTP response (in case there was corresponding HTTP > Request). If no HTTP request is received by the time of the configured > keep-alive time (seconds) the web server closes the connection. Similar opinion from another: [keepalive - How to adjust nginx keepalive\_timeout? - Server Fault](http://serverfault.com/a/331764) > There isn’t a good one-size-fits-all answer; if most client browsers > are able to load all the resources on a page in 5 seconds then 5 > seconds is just fine […] So according to these people the directive is to be calculated in such a way that in that given time (on an average) a web page on your site and its contents (HTML, CSS, JS, images) are completely loaded on most clients. THE QUESTION: So which one is it? (1) or (2)? Who’s right? ---
**Author:** ![Aahan\_Krish](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/a/e47774/32.png) [@Aahan\_Krish](https://www.ruby-forum.com/u/Aahan_Krish)\ **Post date:** [June 17, 2016, 6:28pm UTC](https://www.ruby-forum.com/t/what-exactly-does-keepalive-timeout-work/244801/2 "2016-06-17T18:28:47Z")
You generally want as long a keepalive timeout as you’re able to tolerate. Closing the connection after 5 seconds for example means a full new TCP and TLS handshake has to occur if the user clicks a link after 5 seconds, resulting in a minimum of two RTTs before content, causing a slow loading experience. If you have long pages that could take minutes to read before a user loads another page, then for best performance you want keepalive at 5+ minutes for example. Keep in mind this is an upper limit - the browser is free to choose when to close the connection on the client too. The reason you will sometimes see recommendations to have a low keepalive timeout is to improve performance of legacy forking web servers where each client connection consumed an entire process and thus lots of memory. Having lots of idle clients on such servers caused heavy memory usage and was not recommended. Of course nginx has no such problem since it’s event based. ---
**Author:** ![Aahan\_Krish](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/a/e47774/32.png) [@Aahan\_Krish](https://www.ruby-forum.com/u/Aahan_Krish)\ **Post date:** [June 18, 2016, 1:04pm UTC](https://www.ruby-forum.com/t/what-exactly-does-keepalive-timeout-work/244801/3 "2016-06-18T13:04:30Z")
Hi Richard, thank you for your reply. Since posting on this mailing-list I’ve come across some very good (and credible) descriptions of Keep-Alive, and the best thing is that they are all in agreement, i.e. no conflicting views. I’ll quote them here as a note to self and for everyone; they explain it all. ((1)) [https://tweaked.io/guide/general/webservers/](https://tweaked.io/guide/general/webservers/) ""The purpose of keep-alive is simple in theory; rather than clients having to open a fresh connection to your server for each request they make (index, then CSS, then images), they open one connection. When the initial connection is made that socket is kept open, rather than beign closed, such that further requests can be using it. From the client point of view this improves things, as rather than making the overhead to establish, use, and close, multiple connections only one is used. However the server is left keeping sockets open in the hope that further requests will come, and if they don’t resources are being needlessly consumed which could be better spent on handling fresh visitors. Generally people suggest leaving a small number of sockets available for keep-alive, or only keeping sockets open for a short period of time - such as five seconds - after which time the chances of a further request are minimal.“” ((2)) [https://support.microsoft.com/en-us/kb/813827](https://support.microsoft.com/en-us/kb/813827) ""When Internet Explorer establishes a persistent HTTP connection with a Web server (by using Connection: Keep-Alive headers), Internet Explorer reuses the same TCP/IP socket that was used to receive the initial request until the socket is idle for one minute. After the connection is idle for one minute, Internet Explorer resets the connection. A new TCP/IP socket is used to receive additional requests. You may want to change the HTTP KeepAliveTimeout value in Internet Explorer. If either the client browser (Internet Explorer) or the Web server has a lower KeepAlive value, it is the limiting factor. For example, if the client has a two-minute timeout, and the Web server has a one-minute timeout, the maximum timeout is one minute. Either the client or the server can be the limiting factor. By default, Internet Explorer has a KeepAliveTimeout value of one minute and an additional limiting factor (ServerInfoTimeout) of two minutes. Either setting can cause Internet Explorer to reset the socket.“” ((3)) [http://kb.mozillazine.org/Network.http.keep-alive.timeout](http://kb.mozillazine.org/Network.http.keep-alive.timeout) ""Amount of time in seconds to keep keep-alive connections alive. Default: 115 seconds. Setting this to more than 115 probably won’t help and will make things worse.“” NOTE: More on this in ((5)). ((4)) [http://kb.mozillazine.org/Network.http.max-persistent-connections-per-server](http://kb.mozillazine.org/Network.http.max-persistent-connections-per-server) On the maximum number of HTTP keep-alive connections the application can have open at once to a single server: “Anything above 10 is excessive.” ((5)) [https://blog.fastmail.com/2011/06/28/http-keep-alive-connection-timeouts/](https://blog.fastmail.com/2011/06/28/http-keep-alive-connection-timeouts/) NOTE: That blog post is a must-read and also states when one would want to have long HTTP keep-alive timeout on their server. After following 4/5 links above, I have no questions on how the `keepalive_timeout` directive works or what it’s intended for. I’m satisfied!