Bug 20160

Upon restarting Apache for an upgrade, I noticed that all of the child threads
were going defunct immediately after beginning and load promptly soared to
unheard of levels on our production server.  It turns out our
/var/log/www/access_log was at 2GB so log writes were failing and the child
thread would die out.  /var/log/www/error_log had plenty of space left, but the
child thread did not write there.  It just died.  It had no fatal errors, it
just died.  Do you see the problem?  This is extremely poor error handling --
this is a webserver, not a compiler.  The number one job is to stay on-line and
performing.  Everything else is secondary, including whether or not the log
service is functioning.