On 19/02/16, Christopher Schultz wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Oliver, > > On 2/19/16 10:11 AM, Oliver Graute wrote: > > On 19/02/16, Aurélien Terrestris wrote: > >> Hello, > >> > >> I'm not sure we can accept only one connection at a given time. I > >> tested with the "prefork MPM", and I only achieve 1 concurrent > >> request being processed at a given time and all others requests > >> are buffered (ListenBackLog doesn't seem to work on my server). > >> Once the 1st request has been processed, the other ones are > >> processed one after the other. > >> > >> If running on a Linux, maybe should you configure the iptables to > >> limit connections to 1 for the httpd port. Behind a F5 > >> loadbalancer, there is the Connection Limit parameter on the > >> virtual server which can be set. > >> > >> > >> my httpd conf : > >> > >> ./configure ..... --with-mpm=prefork > >> > >> <IfModule mpm_prefork_module> StartServers 1 > >> MinSpareServers 1 MaxSpareServers 0 > >> MaxRequestWorkers 1 MaxConnectionsPerChild 0 > >> ListenBacklog 1 ServerLimit 1 > >> </IfModule> > > > > thx for this proposal, i will try it this way... > > I'm kind of curious why you are using Apache httpd if you just want a > single connection to be available. Anyone can write a > single-connection HTTP service in about two pages of higher-level > source code (e.g. Python, Java, C#, maybe even Perl). This is a bit > like using a Saturn V rocket to launch a weather balloon. I know that using apache is a bit curious for a embedded device. But today the performance of such an device is like a PC 10 years ago. The costumer needs features like php, mod_nss, json on top of this webserver. Apache met all of these and using apache is a requirement to. > For a single connection, make sure that you disable keepalive -- or at > least set the keep alive timeout to something small, like 1 second. > Otherwise, you'll DOS yourself quite easily. I would also set the TCP > backlog to something small. Maybe as little as 1 (or 0, if that > doesn't actually mean "backlog until we run out of memory"). ok i'll try Best regards, Oliver --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|