It's Oracle RAC, so yes, a cluster.

Yes, I'm REALLY sure it's single threaded.

Yes, no way for someone else to change those rows, since we audit the tables 
with a trigger, and they weren't modified by anything else. 

The data type of the trans_id column is Number (12,0)

Ken

On Sep 24, 2010, at 12:37 PM, Miguel Arroz wrote:

> Hi!
> 
>  What kind of DB are you using? Is it a single server, or a cluster?
> 
>  Are you REALLY sure it's single threaded (although that should not make a 
> difference, because OL doesn't work anyway ;) )?
> 
>  Are you REALLY sure no other app, person, alien, cosmic ray, etc, is 
> changing the conflicting rows at the same time?
> 
>  (Not my suggestion, but really good point ;) ) What is the data type of the 
> column you use for locking?
> 
>  Regards,
> 
> Miguel Arroz
> 
> On 2010/09/24, at 16:56, Ken Anderson wrote:
> 
>> All,
>> 
>> I have an odd problem I'm wondering if anyone else has seen before.
>> 
>> I have apps that do high throughput processing of data - many times a 
>> second.  It is single threaded, and uses a single EOF stack.  I'm sure it's 
>> single threaded because the requests are coming in via an inbound queue, not 
>> any kind of front end.
>> 
>> Every once in a while under significantly high load, I get a few optimistic 
>> lock exceptions in a row, on rows that are only being touched by this app.  
>> We audit every update, and I can look into the audit tables and verify that 
>> nothing else has modified the record except this app.
>> 
>> It's almost like the snapshot has not been recorded properly before the next 
>> request is processed, so EOF thinks something else updated the value.
>> 
>> Our locking is implemented on a single column, trans_id, which is updated 
>> with every save.  The audit table also saves the trans_id that's responsible 
>> for moving the record into audit, and all the values match in succession.
>> 
>> Has anyone had anything like this happen?  Running 5.4.3 on Linux.  and 
>> no... no Wonder.
>> 
>> Ken _______________________________________________
>> Do not post admin requests to the list. They will be ignored.
>> Webobjects-dev mailing list      ([email protected])
>> Help/Unsubscribe/Update your Subscription:
>> http://lists.apple.com/mailman/options/webobjects-dev/arroz%40guiamac.com
>> 
>> This email sent to [email protected]
> 
> 

 _______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com

This email sent to [email protected]

Reply via email to