On Tue, Aug 02, 2005 at 02:08:47AM +0800, Bo Yang wrote:
> Neil, how do scgi_server deal with time-out?
> say, one child process after another get stuck somewhere and never
> return their requests?

The Python code in the scgi package is designed to not need a
timeout.  It is the application's responsibility to not get stuck
itself.

> is there a way to set a timeout somewhere to kill them off and spawn
> new replacements?

If the application process dies then the scgi master process will
respawn a new one.  You could use signal.alarm().  Personally, I
would try to figure out why the application is getting permanently
suck.  That sounds like a bug, maybe in the MySQL extension.

You might try giving one of your suck processes a SIGINT signal.
That may produce a traceback from Python that would help in
determining where the problem lies.

> (I'm experimenting with lighttpd. so far my problem persists, so i
> don't think it's mod_scgi/apache2 related)

Yes, it seems unlikely that mod_scgi is responsible for the problem.

  Neil