It was intended to be sent to the list I guess :) oh, well... ---------- Forwarded message ---------- Date: Thu, 31 Aug 2000 07:26:29 -0500 From: George Sanderson <[EMAIL PROTECTED]> To: Stas Bekman <[EMAIL PROTECTED]> Subject: Re: SIGTERM/SIGKILL at the stop/restart events At 11:18 AM 8/31/00 +0200, you wrote: >I'm documenting the PERL_DESTRUCT_LEVEL options, which skips the >perl_destruct() call. At the same place I also mention that whe you >stop/restart Apache, the parent first sends the SIGTERM (nice) kill signal >to the children, advising them to quit. > >So the parent waits for a few secs and then becomes unpatient and sends >the cruel SIGKILL saying: > >There is nothing you can do against the cruel sysadmin so there comes >a last cry: > >and voila the processes has been killed. > >And one of the nice folks has pointed out that it doesn't matter whether >you have the PERL_DESTRUCT_LEVEL option set to -1, if the processes get >brutally killed, they will not complete their destroy/end blocks and >therefore nasty things might happen. > >Anyone can comment on this possible problem? I've seen many times the >Apache processes being killed with kill -9 (SIGKILL), but I had never had >a significant cleanup to do. Do you? > >Because if there is a problem even a potential, it should be fixed or >prevented fron happening I think. > >Thanks a lot! > I am also mystified by the whole Apache/mod_perl start/stop process. If someone has any good references or comments about this subject, please, provide them. I have a root httpd SIGHUP issue, where an Apache mod_perl module (Apache::Icon) fails to reread directives from the httpd.conf file. It's as if the module gets removed but never re- added after a SIGHUP, therefore the root httpd process also dies (no joy:(. I hope you don't mind my removing the theatrics form the above message (if so, sorry).