Re: [Openflow-dev] Zcatalog Workitem indexing fault

[<bor...@ya...>]

 Hello Thanks a lot to Daniele for his instant reply ! I've downloaded the patch, created from scratch a new openflow folder and a process. Now it's really better but seems the problem is not entirely solved. I think I have to explain in short how I use Openflow: I had the necessity to control a quite complex flow between automatic applications (running on different machines) to be executed in sequence or in parallel depending also on some conditions. Because Openflow does not directly support calls for external urls (out of Zope I mean) and anyway in that case was necessary  to develop url-web enabled applications (like servlets), we had the idea to use Openflow as a Job dispenser and a queue manager.We have therefore several client side applications that poll the Openflow engine asking for a job on a specific activity, Openflow replies with an empty xml if there is no job or with a xml containing relevant data otherwise. I have written a script for this task, it basically search by mean of catalog an inactive workitem, activate it using OpenflowAPI activateWorkitem and generates the xml. Obviously the external applications after receiving the xml do their job and after completion direcly call OpenflowAPI completeWorkitem. Well, I discovered the most critical point is the polling of several applications to Openflow, namely on the script mentioned before. In fact I've implemented the polling mechanism using linux crontabs and in a 1st version every say 2 minutes I invoche 5 or 6 times the same script (on different activities). As a result I saw on the Zope log, messages like:  2003-02-19T09:03:03 INFO(0) ZODB conflict error at /openflow/mywf/start1stavailableJob (611 conflicts since startup at 2003-02-18T15:04:57)  this seem not to be a real problem but I think is connected to these others:  2003-02-18T16:28:03 ERROR(200) FieldIndex unindex_object could not remove documentId -271912924 from index priority.  This should not happen.
Traceback (innermost last):
  File /home/openflow/Zope-2.6.1-linux2-x86/lib/python/Products/PluginIndexes/common/UnIndex.py, line 172, in removeForwardIndexEntry
    (Object: priority)
  File /home/openflow/Zope-2.6.1-linux2-x86/lib/python/ZODB/Connection.py, line 525, in setstate
ReadConflictError: database read conflict error (oid 00000000000098f4, class BTrees._IIBTree.IISet)
------
2003-02-18T16:28:03 ERROR(200) FieldIndex unindex_object could not remove documentId -271912924 from index customer.  This should not happen.
Traceback (innermost last):
  File /home/openflow/Zope-2.6.1-linux2-x86/lib/python/Products/PluginIndexes/common/UnIndex.py, line 172, in removeForwardIndexEntry
    (Object: customer)
  File /home/openflow/Zope-2.6.1-linux2-x86/lib/python/ZODB/Connection.py, line 525, in setstate
ReadConflictError: database read conflict error (oid 00000000000098f6, class BTrees._IIBTree.IISet)  As a result, again I get some NOT INDEXED workitems in Zcatalog.  I tried also to spread application calls to Openflow in order to avoid conflict and seem also the main problem is avoided in this way. My personal opinion is that trere are some problem with concurrent access to instance objects ...   I hope to heve been clear   Thanks a lot in advance bye Roberto
 Daniele <d.t...@ic...> wrote:
Thanks Roberto for the error report.

The bug has been fixed in the cvs. The bug fix will be integrated in the next 
OpenFlow release.

CVS repository:
http://sourceforge.net/cvs/?group_id=2370

Daniele =)

On Tuesday 18 February 2003 09:03, Roberto Borgotallo wrote:
> Hello all
>
> I'm working at a project using Openlow and I recently have discovered a
> problem related, I think to the Zcatalog class.
>
> After adding several instances I found that during the forward phase some
> workitems are created but not indexed. Looking at the Zope log I saw 

 

...



---------------------------------
Yahoo! Cellulari: loghi, suonerie, picture message per il tuo telefonino