Re: [PHP4BETA] Re: yet another Session oddness/bug?

> Ah, I hoped to see some kind of error message. > > Your earlier email contained "Failed to write session data" messages. This > indicates that the session file could not be opened. The additional error > checks will print more useful information. > > > here is my script again > > The script looks ok. > > Make sure that your server OS isn't sabotating you. I played with OpenBSD > a little bit and the generic kernel blocks TCP connections, because some > kernel map was full after a few thousand requests. > > Can you please try to reproduce the bug in the following way: > > Download http_load from http://www.acme.com/software/http_load > > Create a file called url containing this: > > http://yourserver/yourscript.php?PHPSESSID=justatest > > Run http_load -parallel 10 -seconds 600 url 2>/dev/null > > The program will output the number of fetches. Exactly the same number should > be in /tmp/sess_justatest. > > - Sascha Ok I just did the test that you described above using http_load. I actually already had it because I was looking for existing test tools for this kind of stuff, before I decided to write my own :) Anyways, so I ran http_load on one box hitting the same exact php script I've been using in all of my examples. Here are the results http_load: [waltb@Hans http_load]$ ./http_load -parallel 10 -seconds 300 ass 2>/dev/null 53565 fetches, 10 max parallel, 2.15975e+06 bytes, in 300.002 seconds 40.3201 mean bytes/connection 178.549 fetches/sec, 7199.11 bytes/sec 0.433859 mean msecs/connect 55.362 mean msecs/first-response 53563 bad byte counts the sess_justatest file: [waltb@hemna /tmp]$ cat sess_justatest2 count|i:53574; As you can see there is a difference between the # of fetches and count in the session file. Walt