Hash: SHA1

Perry wrote:
> Dieter Maurer schrieb:
>> Joachim Schmitz wrote at 2007-6-16 13:35 +0200:
>>> ...
>>> during our recent search for the cause of the many conflict errors, we 
>>> discovered that nearly all of them where caused during indexing the 
>>> portal_catalog, when the same values are indexed. This can happen easily 
>>> with date_indexes, since the dates are only stored with a resolution of 
>>> 1 minute.
>> Usually, this will not cause a conflict -- only when in addition
>> the first objects that use a new date value are created
>> in concurrent transactions. I expect that the probability is not
>> too high for this....
> That was certainly the case with the date_indexes. We could get rid of 
> those conflict errors by using the QueueCatalog. But still we are 
> getting conflict errors on the indexes, we are updating immediatly for 
> example the review_state.
>>> Even more conflict error prone is the reviewstate. Since there 
>>> are only a few reviewstates.
>> Again, only the initial "filling" of the document list for
>> a given state is critical. Once there is a document list,
>> this list (an "IITreeSet", usually) can concurrently add elements
>> (using application specific conflict resolution -- reducing the
>> conflict probability by a factor of about 120).
> I didn't find any application specific conflict resolution in the 
> fieldindex ? Are you saying, that once there is one document created 
> within a review_state, conflict errors for that review_state are very 
> unlikely, even under heavy load.

FieldIndex uses persistent subobjects, mostly BTree variants, to manage
its data structures;  those objects, once populated, supply the conflict

- --
Tres Seaver          +1 540-429-0999          [EMAIL PROTECTED]
Palladion Software   "Excellence by Design"    http://palladion.com
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


Zope-CMF maillist  -  Zope-CMF@lists.zope.org

See http://collector.zope.org/CMF for bug reports and feature requests

Reply via email to