Received: by taz.hyperreal.com (8.7.5/V2.0) id VAA22965;
 Mon, 16 Sep 1996 21:41:53 -0700 (PDT)
Received: from localhost by taz.hyperreal.com (8.7.5/V2.0) with SMTP id
 VAA22957; Mon, 16 Sep 1996 21:41:50 -0700 (PDT)
Date: Mon, 16 Sep 1996 21:41:50 -0700 (PDT)
From: Brian Behlendorf <brian@hyperreal.com>
To: new-httpd@hyperreal.com
Subject: Taming the server if it cannot fork (fwd)
Message-ID: <Pine.BSI.3.95.960916214142.22706F-100000@taz.hyperreal.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-new-httpd@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com



---------- Forwarded message ----------
Date: Mon, 16 Sep 1996 09:40:32 +0200
From: Andreas Koenig <k@anna.in-berlin.de>
Reply-To: andreas.koenig@franz.ww.tu-berlin.de
To: new-httpd@hyperreal.com
Subject: Taming the server if it cannot fork

I recently watched the 1.1.1 server in a situation where it could not
fork all of its children (for reasons you do not want to know). I
discovered that make_child fails too quickly in such a case so the
loop around the make_child() call consumes the whole CPU. I'd suggest
that make_child sleeps for a second if it fails to fork. That way you
reduce system load in a situation that is most probably already
unpleasant.

Here's my proposal:

--- http_main.c-1.63  Mon Sep 16 09:27:34 1996
+++ http_main.c Mon Sep 16 09:27:43 1996
@@ -1549,6 +1549,7 @@
 
     if ((pid = fork()) == -1) {
        log_unixerr("fork", NULL, "Unable to fork new process", server_conf);
+       sleep 1;
        return;
     } 
     


Regards,
andreas