Are you doing testing from another machine? Because of the large numbers of file descriptors open, its advisable to do testing on another machine.
Also, make sure to set your ulimit to a high value. Check your syslog for messages like the following: socket: Too many open files (24) which indicates the system has run out of available sockets (file descriptors). Johnathan On Thu, Nov 18, 2010 at 12:53 AM, Sudeep S <[email protected]> wrote: > Dear APE Team, > > I am using the functionality of inlinepush for my application. We > are > expecting 50K users at any time to be using our application. > > I was doing a load testing with 10K users initially for the same. > For testing I have written a client application which spawns 10K > threads.Each thread does the foll which is exactly how our application > will work > 1) Connects to the APE server on > http://0.ape > .<domain>.com:80?[{"cmd":"CONNECT","chl":1,"params":{"param1":"value1","param2":"value2","param3":"value3","param4":"value4","param5":"value5"}}]] > 2) Validate the above parameters at the server end and > authenticating them by sending the request to another URL (using > Http.js on server end for the same) > 3) On authentication each thread creates a new channel. > 4) Each thread then connects to ape server and checks if any > message > is there on its channel or not. After every 25s the each thread sends > another request to check the same. > > We are sending messages once every 2 hours (In real application the > frequency will be less every 2 mins or so and we will be using a > script similar to the Php bundled with inline push demo).All the > thread are able to receive their own messages properly. > Our load testing runs works smoothly for 19-20 approx hours with 10K > users and all the channels are created. But after that the APE server > stops responding, surprisingly when I check the process listing, it > shows me that APE is still running and also the PID is present in / > proc file. > During peak times of load testing, we are having 14-15 K http > connections out of which 13-14K are in ESTABLISHED state. > > My server configruations are as follows > OS:- CentOS release 5.4 (Final) 32 bit > Kernel:- Linux 2.6.18-164.el5PAE > APE Version:- 1.0 > Server Side JS:- mootools.js; Http.js; utils.js; inlinepush.js > Also I have came across a performance tuning post in the google > groups specifying tuning of sysctl and other TCP based parameters. > > We are currently using all default values for the same. It > would be really great if you can share some sysctl parameters using > which I can ramp up my testing scenario till 50K users, We wish to > keep the tests running for maximum 24-30 hours. > > Thanking You, > > Regards, > Sudeep Sunthankar > > -- > You received this message because you are subscribed to the Google > Groups "APE Project" group. > To post to this group, send email to [email protected] > To unsubscribe from this group, send email to > [email protected]<ape-project%[email protected]> > For more options, visit this group at > http://groups.google.com/group/ape-project?hl=en > --- > APE Project (Ajax Push Engine) > Official website : http://www.ape-project.org/ > Git Hub : http://github.com/APE-Project/ > -- You received this message because you are subscribed to the Google Groups "APE Project" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/ape-project?hl=en --- APE Project (Ajax Push Engine) Official website : http://www.ape-project.org/ Git Hub : http://github.com/APE-Project/
