[Zope-DB] Problems with Oracle DA and Dates

Matthew T. Kromer matt@zope.com
Mon, 01 Oct 2001 09:43:35 -0400


Dario Lopez-K=E4sten wrote:

>(I apologise in advance for the crosspost, but I think this is a valid
>question on both the zope-db and zope-dev lists. If you disagree, flame
>away, and I'll never do it again. oh, btw: flame in private mail, please=
)
>
>Hello!
>
>We have run into a showstopper problem here where it seems (we're not su=
re
>yet) that there is a severe problem using dates returned from the Oracle=
 DA
>adapter. Other possible culprits include LocalFS, Transparent Folder,
>Formulator and the source release of Zope 2.4.1.
>
>The problem is that Zope either dies, core dumps and dies, or slows down=
 to
>a crawl.
>
>We are using the Zope 2.4.1 release, with Transparent folders and LocalF=
S,
>latest, and a sligthly modified Formulator.
>
>There are about 2-6 people working and developing in it during all hours=
 of
>the day (24 hours).
>
>Unfortunately nothing shows up in any of the logs, so they are of little
>use; I don't even have a traceback so display. We *think* we found somet=
ign
>pointing at LocalFS in one of the coredumps, but we are lowly non-unix
>programmers, and have no idea if this is accurate info or not. It could =
just
>be un-collected garabage memory.
>
>Is anybody noticing anything similar, or if you have any opinion on what
>might be going on, please reply; we are in DS mode here (we are having a
>prototype presentation during two weeks, starting tomorrow) and are feel=
ing
>a bit desperate.
>
>
>Sincerely,
>
>/dario
>

(Replying only to Zope-DB :-D )

Hi Dario, have you looked at the Control Panel with the Debug=20
Information?  It will show you the top refcounted Zope objects; this may=20
help you determine for sure that it is DateTime objects being leaked, or=20
something else.  I dont know if you're using DCOracle2 or not -- I=20
*know* there is a bad bug in DCOracle2 that sneaked in sometime before=20
beta  5 where it leaks data for every request (there's a spot where the=20
result set allocation is duplicated, about a screen apart).  The CVS=20
source for this is fixed; and I've also built Solaris binaries, but=20
haven't made windows binaries yet so I haven't put out a new release.