Taavi wrote:
> Hmm, it shouldn't give any warnings, unless you have multiple
> virtualhosts on the same ip/port (you can't use name-based virtualhosts
> with ssl anyway).

> On my CentOS 5 box, if I set NameVirtualhost on a ip/port that's used
> for SSL, I basically get the same thing. Checking it with wireshark
> shows that the browser and server start communicating over ssl, but then
> the server randomly responds with a plain html error (this probably
> creates the "ssl_error_rx_record_too_long" error you see on the browser
> because it tries to parse this as ssl). Not sure why it does that, but
> removing the Namevirtualhost on that port seemed to resolve this issue.

Well, I've had an interesting turn of events. I'm not quite sure why
its started working again. I spent a good few days trying to get it
working with mod_ssl, but it wasn't working right as described - so I
started using pound as a temporary resolve. After your post, I tried
again with this in my apache virtualhost conf.

<VirtualHost 1.2.3.4:443>
  ServerName www.example.co.uk
  ServerEnvironment example

  #MinSpareProcessors 4
  MaxProcessors 10

  DocumentRoot /domains/example.co.uk/http/
  ServerAlias *.example.co.uk
  ServerAlias example.co.uk

  SSLEngine On
  SSLCertificateFile /domains/example.co.uk/ssl_keys/example.crt
  SSLCertificateKeyFile /domains/example.co.uk/ssl_keys/example.key
  #SetEnvIf User-Agent ..*MSIE.*. nokeepalive ssl-unclean-shutdown

</VirtualHost>


And now it works perfect! I have no idea why, but it is working
perfect again, so pound has been stripped off (for now).

> Would it be possible that you compile peruser with debugging enabled (by
> uncommenting the MPM_PERUSER_DEBUG in peruser.c) and attach the debug
> log about these seg faulting children? Note that the debug creates
> pretty much IO load on the server, so I wouldn't suggest running it for
> along time.

Again, I have no idea why, but I tried again this morning, put the
particular user in question back into a chroot to test it - and it
works fine! I am beyond confused! So, fingers crossed, everything is
working well.

We have 1 bug remaining now and it stems from when apache is
restarted, it doesn't do it every time, its more at random, but around
20% of all restarts will fail to come back up again. Here's 5 lines of
output around the error from the apache2 log. Its obviously something
to do with the NumServers being set - but as far as I can see, this
isn't a setting I can change in peruser? We have 52 server
environments on the server. The closest thing I could find was,

ServerLimit 300
MaxClients 150

Here is the log output:

11555365-sh: host: command not found
11555393-sh: host: command not found
11555421-[Thu Jun 04 19:59:40 2009] [notice] SIGHUP received.
Attempting to restart
11555562-Syntax error on line 250 of /etc/apache2/peruser.conf:
11555617:Trying to use more server environments than NumServers.
Increase NumServers in your config file.
11555715-[Thu Jun 04 19:59:44 2009] [info] Init: Seeding PRNG with 656
bytes of entropy
11555794-[Thu Jun 04 19:59:44 2009] [info] Init: Generating temporary
RSA private keys (512/1024 bits)
11555888-[Thu Jun 04 19:59:44 2009] [info] Init: Generating temporary
DH parameters (512/1024 bits)
11555979-[Thu Jun 04 19:59:44 2009] [info] Init: Initializing
(virtual) servers for SSL
11556058-[Thu Jun 04 19:59:44 2009] [info] mod_ssl/2.2.9 compiled
against Server: Apache/2.2.9, Library: OpenSSL/0.9.8g

I've changed the serverlimit to 500 to see if it has any effect.
_______________________________________________
Peruser mailing list
[email protected]
http://www.telana.com/mailman/listinfo/peruser

Reply via email to