[Zope-dev] Zope 2.4.4b1 dumps core

Dario Lopez-Kästen dario@ita.chalmers.se
Thu, 14 Feb 2002 08:48:53 +0100


----- Original Message -----
From: "Andreas Jung" <andreas@zope.com>
To: "Dario Lopez-Kästen" <dario@ita.chalmers.se>; <zope-dev@zope.org>
Sent: Wednesday, February 13, 2002 7:52 PM
Subject: Re: [Zope-dev] Zope 2.4.4b1 dumps core


> Does this problem persist when you remove the 3rd-party products ?
> Are you running Zope with enabled garbage collector ?
>
> - aj
>

Hm... I think I am - I have not done anything to turn it off.

Unfortuntatly I can't remove all 3rd party products - if I do, I don't have
much of a testcase, because then I don't have an app to test against. Like I
remove DCO2 I only get errors, because lots of data is used in presenting
layouts, pages etc.

If I remove formulator, I get errors, because the pages that I would need to
test, give errors.

I can remove TransparentFolders and possibly strippogram, but we have
previously done tests to see if TP slowed things down, but got no colcusive
results (and no decrease in core dumps either).

I can also remove Strippogram, but it is a late addition. We have had core
dumps since the days of borked Python 2.1 and Zope 2.4.3/2.4.2 - culprits
have been, in the past, old versions of DCO2, previous versions of Python
and old versions of exuserfolder (which we don't use anymore).

The only thing I have to go on is that as son as there are lots of
PythonScripts involved thins star to deteriorate.

I might also add that changing zope to use only one thread, as has been
mentioned as a possible workaround with pesky DA's, is not a realistic
option - on the contrary I need to bump up the thread count to around 10, to
be able to use more than 4 concurrent SQL queries.

I am at loss here - I don't even know where to begin looking for errors.
Maybe it's our app that is faulty (I know for a fact that around 40% of it
could be done in a cleaner way).

Oh, and the icing on the cake is that allmost all of my PythonScripts need
to be recompiled, all of a sudden. All I did was to pack the ZODB, and copy
it to production environment. The app is only about 6 megs in zie all in
all, so I wouldn't expect any serious ZODB corruption (as the bulk of the
data.fs is moslty app-logic, I would expect corruption to show itself real
fast, if it existed).

So, apart from killing myself, is there a way out of this? Or at least a
general direction in which to start looking for possible solutions?

I have to deploy this app 2 days ago, so I 'll try to setup as many
safegueards as I can. After that, in a sanndbox, I'll set my app up with as
few 3rd party extensions as possible, and see if it helps.

Any insight is greatly appreciated. I have saved coredumps if anyone would
care to dig thru 20-40 megs of data :-).

Sincerely,

/dario