On Tuesday 22 September 2009 3:04:46 pm Jeff Trawick wrote:
> On Tue, Sep 22, 2009 at 4:50 PM, Ricardo Cantu <[email protected]> wrote:
> > Okay, got it all compiled and working. I've stressed here and there and
> > seems
> > that it is running great! I see a little performance boost over the
> > mod_fastcgi module as well.
> >
> > So far the only issue I've come across is when my C program is exited it
> > leaves a zombie process for a couple of seconds then get's cleaned up.
> > Log confirms this as well;
> >
> > [Tue Sep 22 14:40:57 2009] [notice] mod_fcgid: process
> > /var/www/html/engine/bin/engine_10(12453) exit(normal exit), terminated
> > by calling exit(), return code: 0
> > [Tue Sep 22 14:40:57 2009] [warn] mod_fcgid: cleanup zombie process 12453
> >
> > Is this normal? or is there a directive that would help it exit cleaner?
>
> Set FCGIDZombieScanInterval, which defaults to 3 seconds, down to 1. But
> note that it may not scan for zombies every second because of this <=
> comparison:
>
> /* Should I check zombie processes in idle list now? */
> if (procmgr_must_exit()
>
> || apr_time_sec(now) - apr_time_sec(lastzombiescan) <=
>
> sconf->zombie_scan_interval)
> return;
>
> Maybe set it to zero to mean 1 second ;) It doesn't currently complain
> about a setting of 0.
>
> (It would be nice not to have to scan for zombies at a set interval. The
> process manager could be awakened when a subprocess exits, at least on
> Unix.)
>
Okay, so this looks like the place to register some type of "wake up" callback
on exit.
fcgid_proc_unix.c
apr_status_t
proc_spawn_process(char *lpszwapper, fcgid_proc_info * procinfo,
fcgid_procnode * procnode)
.
.
.
/* Unlink it when process exit */
if (ap_unixd_config.suexec_enabled) {
apr_pool_cleanup_register(procnode->proc_pool,
procnode, socket_file_cleanup,
exec_setuid_cleanup);
} else {
apr_pool_cleanup_register(procnode->proc_pool,
procnode, socket_file_cleanup,
apr_pool_cleanup_null);
}
.
.
.
And this is what we would want to wake up:
fcgid_pm_main.c
apr_status_t pm_main(server_rec * main_server, apr_pool_t * configpool)
{
.
.
/* Wait for command */
if (procmgr_peek_cmd(&command, main_server) == APR_SUCCESS) {
if (is_spawn_allowed(main_server, &command))
fastcgi_spawn(&command, main_server, configpool);
procmgr_finish_notify(main_server);
}
.
.
.
What you think?
