[Zope] Degrading performance with redirects?

Thomas B. Passin tpassin@mitretek.org
Mon, 3 Dec 2001 15:37:03 -0500


[Dieter Maurer]

> Jonathan \(Listserv\) writes:
>  > We're experiencing a problem with a Zope based site that serves most of
its
>  > content via redirects, the new location based on results from a
PostgreSQL
>  > database.
>  >
>  > The weird thing is that the Zope site becomes very slow after a week or
two,
>  > which is fixed by restarting Zope. It doesn't seem to be a memory
problem
>  > (512MB free out of 768MB total) or a load problem (server is 99% idle
most
>  > of the time). Zope has about 5 Gigabytes of storage available, running
on
>  > Red Hat Linux 7.1.
>  >
>  > The homepage is served by a redirect:
>  >
>  > <dtml-call "RESPONSE.setHeader('content-type', 'text/html')">
>  > <dtml-call "RESPONSE.redirect('nl/another.page')">
>  >
>  > The rest of the site is basic DTML, building tables based on results
from
>  > the database with a few if's here and there.
> Strange!
>
>   ... because a redirect is just a normal response with a specific
>   HTTP response code and a location header. Nothing special at all.
>
> Sure, you problem is not the redirect.
>
> Unfortunately, I do not know what it is...

Didn't someone track this down somewhat recently (sorry, I forget who)?
Seems to me that the someone found that this behavior was related to threads
that get stuck trying to get a non-existent response somehow.  Since you
can't kill off a thread that's blocking on a socket, you could get clogged
with stuck threads.  Should be able to find it, it must have been within the
last month on this list.

The right log reports supported this, as I remember.  But why all the stuck
connections, well...

Cheers,

Tom P