Re: puzzling deadlock

"Brian Peasland" <dba_at_remove_spam.peasland.com> wrote in message news:3F730CF2.50877C6A_at_remove_spam.peasland.com...
> You'll have to look more closely at the application's code. It is
> impossible for two concurrent sessions to issue just one DML statement
> and cause a deadlock. There have to be other DML statements involved. So
> you'll have to look at all the DML statements to figure out what is
> getting locked by concurrent sessions leading to a deadlock condition.

Hi!

I don't have the time to test it now, but if two update statements with different execution path fire, they could cause a deadlock, when one update is done using FTS and another one using an index for example:

1. FTS access updates row 1
2. index access updates row 1000000
3. index access then tries to update 1 but remains waiting
4. FTS will eventually reach locked row 1000000 as well

Tanel. Received on Sun Sep 28 2003 - 20:57:06 CDT