[Repoze-dev] CMF problems under wsgiutils server

[Chris McDonough]

I should note that the problem is a little more insidious than this;  
it's usually evident when you're using a browser (like reloading the  
CMF site page from within a browser), so that there are concurrent  
connections.  Serialized connections (such as from ab -n100) don't  
expose the problem.

On Sep 23, 2007, at 1:09 AM, Chris McDonough wrote:

> FYI,
>
> When repoze.zope2 is run using either Paste's wsgiutils server (or  
> the slightly simpler, but still wsgutils-derived server I created  
> in repoze.zope2.server), and when either the transaction manager is  
> configured into the pipeline or repoze.zope2 detects it isn't in  
> the pipeline and closes the ZODB connection itself, every 5-10  
> requests or so fail with a skin lookup error.  If the txn manager  
> isn't in the pipeline and you change repoze.zope2 to prevent it  
> from closing the connection after each request when it detects  
> repoze.tm isn't present, the problem goes away.
>
> The wsgiutils server is based on SimpleHTTPServer, which creates a  
> new thread for each request.
>
> I'm guessing this is due to the interaction between repoze  
> connection management and the way that CMFCore.Skinnable arranges  
> cleanup of its  "skindata" by placing a closer object into the  
> request keyed on the current thread id.
>
> The problem doesn't appear to be evident under mod_wsgi.
>
> - C
>
>
> _______________________________________________
> Repoze-dev mailing list
> Repoze-dev at lists.repoze.org
> http://lists.repoze.org/mailman/listinfo/repoze-dev
>

_______________________________________________
Repoze-dev mailing list
Repoze-dev at lists.repoze.org
http://lists.repoze.org/mailman/listinfo/repoze-dev