Hi,
In my web server, when I run a stress test with many clients, after a long time I suddenly get an error on the accept system call and the error is "Too many open files".
When I do a socklist (or netstat), I see that there are only 400 sockets open (and I have 400 clients, so that is correct).
However when I "cat /proc/PID/fd" I see 1024 files. I think accept failed because of this.
Yes. 1024 is the usual Linux limit on filehandles for a single process. I believe you can change this limit during kernel compilation, but offhand I do not remember how.
My web server is as simple as possible (it just serves up a static string compiled in with the server). I know I am closing the sockets... what else could I be doing wrong?
Depends on what the files are.
Are you really seeing them with "cat /proc/PID/fd"? (This gives me an error.) And not "ls -l /proc/PID/fd"? (This works here.) Whatever. Use the second choice to see what the filehandles listed in the pseudo-directory actually point to, and you (or we, if you report a meaningful subset of that info in a followup here) may be able to offer a suggestion.
Depending on details you haven't told us, you may be doing nothing wrong except underestimating the number of filehandles needed per connection ("client").
- To unsubscribe from this list: send the line "unsubscribe linux-newbie" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.linux-learn.org/faqs
