Start » History » Revision 6
Revision 5 (stbuehler, 2012-06-03 19:36) → Revision 6/10 (stbuehler, 2012-08-11 10:42)
h1. weighttp h2. About @weighttp@ (pronounced @weighty@) is a lightweight and small benchmarking tool for webservers. It was designed to be very fast and easy to use and only supports a tiny fraction of the HTTP protocol in order to be lean and simple. @weighttp@ supports multithreading to make good use of modern CPUs with multiple cores as well as asynchronous i/o for concurrent requests within a single thread. For event handling, weighty relies on "libev":http://software.schmorp.de/pkg/libev.html which fits the design perfectly, being lightweight and fast itself. Thanks to that, weighty supports all modern high-performance event interfaces like epoll or kqueue, that the major OSs provide. h2. Deployment h3. Obtaining Use "git":http://git-scm.com/ to fetch the latest source: <pre> $ git clone git://git.lighttpd.net/weighttp </pre> or download a .tar.gz from http://cgit.lighttpd.net/weighttp/snapshot/weighttp-master.tar.gz h3. Bulding Use "waf":http://code.google.com/p/waf/ (included, needs python) to build: <pre> $ ./waf configure $ ./waf build </pre> (If you have libraries/includes in special paths, try something like this: @LINKFLAGS=-L/opt/local/lib CFLAGS=-I/opt/local/include ./waf configure@) See ./waf --help for available configure options and other commands available. Alternatively, you can just use gcc directly too (but you will have to (un)install it manually: <pre> $ gcc -g2 -O2 src/*.c -o weighttp -lev -lpthread </pre> h3. Installing Depending on your configure arguments, you might need root to (un)install. <pre> $ ./waf install </pre> h3. Removing <pre> $ ./waf uninstall </pre> h2. Usage @weighttp@ uses similar commandline arguments as @apache bench@ (ab): <pre> weighttp <options> <url> -n num number of requests (mandatory) -t num threadcount (default: 1) -c num concurrent clients (default: 1) -k keep alive (default: no) -h show help and exit -v show version and exit </pre> h2. Troubleshooting Benchmarking a webserver can result in a lot of sockets being created, especially if you don't use the -k parameter to reuse existing connections. This can cause your system to run out of TCP port numbers and you will see the following error: <pre> error: connect() failed: Cannot assign requested address (99) </pre> Linux has a hardcoded 60 seconds timeout before it reuses a previously in use port. You can mitigate the problem by tuning 2 kernel parameters: h3. TCP_TW_REUSE This allows reusing of sockets which are in the TIME_WAIT state when it is safe from a protocol perspectiv. <pre> echo 1 > /proc/sys/net/ipv4/tcp_tw_reuse </pre> or add "net.ipv4.tcp_tw_reuse = 1" to your /etc/sysctl.conf to make the change permanent. You need to reload it via _sysctl /etc/sysctl.conf_ h3. TCP_TW_RECYCLE This allows fast recycling of sockets which are in the TIME_WAIT state even if it is _not safe_ from a protocol perspective. You should not use this in a production environment. It can cause issues with loadbalancers and other mayhem. <pre> echo 1 > /proc/sys/net/ipv4/tcp_tw_recycle </pre> or add "net.ipv4.tcp_tw_recycle = 1" to your /etc/sysctl.conf to make the change permanent (bad idea). You need to reload it via _sysctl /etc/sysctl.conf_