[Zope] Problem with running same script at a time

[Pletli Antal]

Thank you Dieter!

Here is the exact scenario, which in I first observed the problem:

I have a external zope python module that communicates via TCP with a
background system written in Java.
I have a background system that communicates with the Zope via URL.

I have a demo user folder in my Zope project with two scripts. The first
script sends a tcp message to the background (with the help of external
module), the second one do something when the background call it via
url.

First the Zope script sends a tcp message to background system, and
waits for the reply. The background receives the message and try to call
the url by http request, then if it is success replies to Zope script's
tcp request.
Here is the problem. The background system cannot access the URL  until
in the first script tcp timeout happen.

When I move the second script into other outer folder, all is OK.

It is very strange, and I'v never read from this problem before.

Best regards,

Antal

-----Original Message-----
From: Dieter Maurer [mailto:dieter at handshake.de] 
Sent: Thursday, September 21, 2006 7:46 PM
To: Pletli Antal
Cc: zope at zope.org
Subject: RE: [Zope] Problem with running same script at a time

Pletli Antal wrote at 2006-9-21 09:12 +0200:
> ...
>I use the same product. 
>This problem is one of the others...

I had to fix a problem in "psycopg":

   During "connect", Python's Global Interpreter Lock (GIL)
   was not released (freezing Zope during connect).

   Usually, this is not a problem as "connect" is quite a fast
   operation.

   We noticed it only when the server for the postgres installation
   was down (then "connect" became a several minute operation
   until the TCP timeout happened).

However, it is unlikely, that this is your problem.

>The behaviour is the same when i
>use long socket communication (as a tcp client), etc.

This, too, should not happen -- and I never observed it (but a colleague
reported something like this but never showed it to me -- and I did not
believe him; or more precisly, I tested his alledgedly observed
behaviour and could not reproduce it...)

>In these cases the
>folder wich contains the scripts be unaccessible.

Unbelievable -- unless your folder access triggers the long socket
communication. In this case, it would not be the old socket
communication that "blocks" but the new one.

Of course, your folder access should not trigger anything else...



--
Dieter