I apologize that my description was so confusing.
I appreciate the feedback and so would like to clarify.

The diagram is how the traffic flows.  So we have:
Step 1 - a Load Balancer that passes traffic to Apache

Step 2 - Apache which uses mod_proxy to obfuscate the URL and proxies traffic to
the Zeo Client

Step 3 - the Zeo Client (residing on the same physical box as Apache) that then
forwards the traffic to the Zeo Server

Step 4- the Zeo Server that runs on its own box and writes to the ZODB.

I thought putting Apache in front of Zope was very common (using VirtualHost
Monster) - guess not. 

As for versions:
FreeBSD 5.3
Python 2.3.5
Zope 2.7

The load on the Zeo Server is minuscule.
The load on the Zeo Client(s) gets very large.

It's the python process on the Zeo Client that grows large and seems to become
unresponsive. Possibly due to the "Transaction blocking" being reported in the
Zeo Severs log.  Strangely disconnecting the Zeo Client from the Server while
the Zeo Client is "locked" rarely gives the ... "disconnected during
transaction" error message that I expected to see more of.

Hope that clears up things for anyone else who'd like to chime in.

Dieter - As for your suggestions. Thanks so much!
I will certainly begin to investigate and pursue each.  They seem very good
things to do - just in general. I appreciate the input. I'll let the list knows
how everything works out.

I do notice that there was no mention of any general (config file) configuration
techniques (increasing threads, cache objects, etc. etc.) --- is configuration
not really that big of a performance boost?

Thanks again for the direction.

Zope maillist  -  Zope@zope.org
**   No cross posts or HTML encoding!  **
(Related lists - 
 http://mail.zope.org/mailman/listinfo/zope-dev )

Reply via email to