September 2002 - Zope-Dev

Database Cursors and Gargabe collection ?
by Romain Slootmaekers 08 Sep '02

08 Sep '02

Yo, we have a problem using garbage collection to clean up open database cursors. Our approach is really simple, but doesn't work. :( we have a method( on a SimpleItem object) getCursor: def getCursor(self): """ doc """ if not hasattr(self,'_v_cursor'): from MySQLdb import connect,escape_string db=connect(...) cursor=db.cursor() self._v_cursor=CursorWrapper(cursor) cursor=self._v_cursor.getCursor() return cursor the cursorwrapper is a class that closes the cursor on GC. class CursorWrapper: def __init__(self,cursor): print "created" self.__cursor=cursor def getCursor(self): print "given out" return self.__cursor def __del__(self): self.__cursor.close() del self.__cursor print "deleted" So in theory, the cursor will be closed on GC. In practice however, we get multiple 'created' a lot of 'given out' but no 'deleted' messages in the cucu debugging. After searching the web/zope site the only related thing we could find was a quote from Jim: "old objects don't die, they just fade away' not very useful. So the questions are: 1) why doesn't the __del__ get called on our wrapper ? 2) do we have to resort to closing and opening database cursors every single time we access the RDB ? TIA, Sloot.

1 0

Mp3 lere..
by HaLLoK Ben 08 Sep '02

08 Sep '02

3mp3.net haber listesinden: Yerli ve yabanci albümler tamamen yenilendi. Artik istediginiz mp3 leri zip olarakta sitemizden indirebilirsiniz. Dosya indirme hizi yaklasik 2.5 kat artti. Boylece kisa zamanda tam albümlere sahip olacaksiniz. Istekleriniz dogrultusunda Top25 yenilendi. Beklemeye gerek yok. Sitemizde gezintiye baslayabilirsiniz. http://www.3mp3.net

1 0

Changes to Interface package/HEAD
by Andreas Jung 07 Sep '02

07 Sep '02

I noticed that the HEAD contains a rewritten Interface package that breaks at least two of my products that depend on the Interface.verify. Are these changes documented anywhere and why is there no compatiblity layer for exisiting code? Andreas --------------------------------------------------------------------- - Andreas Jung http://www.andreas-jung.com - - EMail: andreas at andreas-jung.com - - "Life is too short to (re)write parsers" - ---------------------------------------------------------------------

3 2

Zope-Dev digest, Vol 1 #1821 - 6 msgs
by TGERMAINE＠UPei.CA 07 Sep '02

07 Sep '02

I am away on vacation until Sept 16. If your inquiry is urgent please resend your request to Chris Martin (cmartin(a)upei.ca) and he will look after it. Alternatively, you can call Chris at 566-0604 Tom Germaine Access Services Administrator

1 0

Re: [Zope-dev] SWAP issue
by Ahsan Imam 06 Sep '02

06 Sep '02

I introduced the LeakFinder product. However Friday, Saturday and Sunday are least busiest days on our site. Our main zope server is not swapping at all today. So I cannot replicate what happened yesterday or the day before. When I looked at the debug information section, it showed me the request that I had just made. I guess I will have to wait until a busy day. In an earlier mail I had named two references to classes (DocumentTemplate.DT_Util.Eval,OFS.Image.Image ). Should I be concerned with these? These are always topping the list. (Our site is very image intensive). Thanks >>> Dieter Maurer <dieter(a)handshake.de> 09/06/02 11:57 AM >>> Ahsan Imam writes: > Is there a way catch that non terminating loop? This non-terminating loop is just a theory I have. I could be completely wrong. Are there other things I could look at for clues. When you have indeed a non-terminating loop (and you Zope is still responsive), you can find at least the request that triggered it: Go to "Control_Panel --> Debug information" and look at the bottom of this page. You will see the currently active requests. Dieter _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

1 0

Re: [Zope-dev] SWAP issue
by Ahsan Imam 06 Sep '02

06 Sep '02

The kernel is 2.4.7-10. Zope 2.5 on Red Hat 7.2 python 2.1.3 Here is what I think the problem is: Some object or code was introduced on Tuesday morning. The users are beginners so they could have introduced a never ending loop or something of that sort. I turned on profiling to see if thre was anything in particular consuming a lot of time. A the top of the list was Connection.py. Is there a way catch that non terminating loop? This non-terminating loop is just a theory I have. I could be completely wrong. Are there other things I could look at for clues. >>> Toby Dickenson <tdickenson(a)geminidataloggers.com> 09/05/02 06:46 AM >>> On Thursday 05 Sep 2002 2:58 am, Ahsan Imam wrote: > Hello All, > > I am currently running zope 2.5 and python 2.1.3. Are you actually seeing the zope processes use too much memory? 2.5.x have a ZODB cache mechanism that does not respond well to memory pressure. If your application touches many objects per transaction then memory usage may grow out of control. 2.6 improves this significantly. > The machine has 2 gigs of > rams and is running on Red Hat 7.2. Which kernel? > After starting zope eveything works fine for a while. Slowly the swap > starts to swell up Memory that is not being actively used might as well be moved into swap, freeing the ram for more productive uses. This does not necessarily indicate a problem. > and after a few hours the machine has to be rebooted. > The funny thing is that free shows that there is almost a gig of RAM free > and swap keeps on growing. Are you sure you are interpreting the output from 'free' correctly? It is often misinterpreted. (please post the output from 'free -m' if you are not sure.) > After a while the machine simply hangs. > I have recompiled python and zope. Turned on profiling. > I have looked through the stupid.log. There are no core dumps either. I > have also tried running zope with one thread. > > Does anyone have suggestions. you havent explained why you think this is a zope problem.

2 1

Re: [Zope-dev] SWAP issue
by Ahsan Imam 06 Sep '02

06 Sep '02

I installed the LeakFinder product and two most called references are: 1) DocumentTemplate.DT_Util.Eval 2) OFS.Image.File Do these classes have any prior history? Should I patch them or what other course can I take? Your help is much appreciated. Thanks Ahsan Imam >>> Dieter Maurer <dieter(a)handshake.de> 09/05/02 13:14 PM >>> Ahsan Imam writes: > After starting zope eveything works fine for a while. Slowly the swap starts to swell up and after a few hours the machine has to be rebooted. The funny thing is that free shows that there is almost a gig of RAM free and swap keeps on growing. After a while the machine simply hangs. When it is Zope, then restarting Zope should free your swap again. This way, you can test whether Zope is the culprit and avoid rebooting your computer. When it is Zope, there probably is a resource leak. There is the "LeakFinder" product to localize such leaks. Dieter _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

1 0

Problem with structured-text on RedHat
by Arthur Tomas 06 Sep '02

06 Sep '02

i have a big problem using structured-text on Redhat Linux. I need german and french characters, which are not correctly represented. Each word with special character is ignored, for example: **Köln** ist not translated to: <strong>Köln</strong>. What I get to see is still **Köln**, so it will not be translated. I set LC_ALL evironment-variable to "de_DE" or "fr_FR" or "", I try to start ZOPE calling "start -L "de_DE" - does not help. There are also "locales"-directories on my machine with right files for de_DE, fr_FR and so on ... It works correctly on W2K or NT4 with start.bat -L "", but no chance for RedHat Linux... I tried 2 last Versions of RedHat... Any Ideas? Is there some other settings, which could help? Best Greetings ATOM -- Replay To: arthur.tomas(a)web.de ______________________________________________________________________________ WEB.DE MyPage - Ultimatives Kommunikationstool! Ihre Message sofort online! Domain aenderbar! http://www.das.ist.aber.ne.lustige.sache.ms/

1 0

RE: [Zope-CMF] Performance problem in ZOPE 2.5.1, 'getOwner' in/lib/python/AccessControl/Owned.py
by Arnar Lundesgaard 05 Sep '02

05 Sep '02

> What kind of user folder are you using? We are using the extensible User Folder. ZODB Auth BTree Source Basic Membership Source Authentication Type: cookie-based Credential Cache Timeout: 600 (We have also tried 0 without much success) Negative Credential Cache: 0 memberdata is the standard CMF with a exUserfolder wrapper for it. I have just tested the performance with a normal userfolder with only two users, and the response was instantaneous as it should be. This seems to be the reason for our troubles. :-/ We have about 100k users (many inactive) which we cannot remove, so we would very much appreciate experiences with this number of users and similar setups. Arnar Lundesgaard

1 0

Performance problem in ZOPE 2.5.1, 'getOwner' in /lib/python/AccessControl/Owned.py
by Arnar Lundesgaard 05 Sep '02

05 Sep '02

Hi, as I have written on this list before, we have had serious performance problems on one of our CMF based sites. For a long time we believed that this was mostly due to load, and differences in our application. Unfortunately the more we searched for answers the less we understood. Recently we have converted to a ZEO setup, and that has made it much easier to locate hotspots in the application. One of things we have seen, is in our skinned 'folder_contents', which includes the name of the objects owner. This is fetched by using the 'Creator' method from the DublinCore mixin. This again calls getOwner of 'AccessControl/Owned.py'. On the first site we have set up (using the same application, whose performance currently is quite good) and folder_contents is virtualy instantaneous. Therefore we tried using the default skin. The difference was astonishing, and we are seeing a difference of 20-30 seconds. We started removing functionality, and we quickly found the culprit. In the CVS log for Owned.py it says (around 1.15.10.2): Remove cache from getOwner, since it would have cached both unwrapped and wrapped objects (caching wrapped objects being a bad thing) This change seems to be the difference between our two sites, and it will also affect other parts of our application. For instance when we reindex objects on change, since Creator is both an index and a metadata field. Can we simply change this file? What are the consequences of having or not having the cache? Arnar Lundesgaard ------------------------------------- phone: (+47) 982 38 036 mailto:arnar.lundesgaard(at)creuna.no Creuna as Bryggegata 3 NO-0250 Oslo phone office: (+47) 23 23 88 00 fax: (+47) 23 23 88 50 http://www.creuna.no/

3 2