apache-ssl in woody crashing

[David Wilk <myca@cia-g.com>]

Howdy folks,

apache-ssl failing the reload after logrotate is nothing.  I just had an
out and out crash today.  I've administered serveral apache 1.3.x
servers on debian linux (and other unices) w/o incident, however I'm
having a problem with the latest debian woody apache-ssl-1.3.26.

Recently I had apach-ssl just outright crash without any logrotation
involved.  This is a bit alarming.  here's what
/var/log/apache-ssl/error.log shows:

<snip>
[Tue Apr 15 08:08:51 2003] [error] [client 65.101.127.48] Invalid URI in
request GET (null) HTTP/1.0
Failed to connect to socket: /var/run/gcache_port
connect: Connection refused
apache-ssl: gcacheclient.c:118: OpenServer: Assertion `!"couldn't
connect to socket"' failed.
Failed to connect to socket: /var/run/gcache_port
connect: Connection refused
apache-ssl: gcacheclient.c:118: OpenServer: Assertion `!"couldn't
connect to socket"' failed.
[Tue Apr 15 08:25:35 2003] /usr/lib/apache-ssl/gcache started
<snip>

the first line being normal and the last being normal (after restart).

Now, this is debian woody (3.0), apache-ssl-1.3.26+1.48 configured with
130 virtual hosts each with two log files on kernel 2.4.20-stock, 2GB
RAM.  I believe the 2.4 kernel provides plenty of file handles to
processes, so I can't imagine this is an issue.

Anyone else see this?  I'm wondering:
1. is there a prob with this package?  should I compile from latest
sources instead?
2. is it apache-ssl that's the prob, would apache plust mod-ssl do
better?
3 is it not apache at all, but rather some system resource limitation?
I know apache uses a butt-load of file descriptors (file handles), but
the 2.4 kernel isn't stingy with those....

any ideas would be greatly appreciated.

thanks,
Dave

-- 
*******************************
David Wilk
System Administrator
Community Internet Access, Inc.
admin@cia-g.com