Re: PMON

On Thu, 12 Apr 2007 00:12:35 +0200, Uwe Weber wrote:

> Yes, he has something about dead client detection and sqlnet.expire_time
> but the things the Oracle supporter said about an event that would
> possibly do the same thing without involving the network looked
> promising. But googling web and groups came up empty, so maybe he has
> just confused the sqlnet parameter with an event?

UWe, PMON will wait for the processes to roll back their own transactions. There is no PMON event that will prevent that. What you can do on a Unix/Linux platform is to gently move the corresponding server process, visible in V$PROCESS, to another plane of existence, usually by kill -9 <pid>. When you kill the server process, PMON will roll back the transaction and release any locks or latches that process might have held before passing through the Pearly Gates. You will still have to wait for the transaction rollback to complete. If you were trying to delete 10,000,000 rows from a BFT (very big table, an expression known to any Doom player), you will still have to wait for rollback and locks to be released. There is no other way. Even if you do shutdown abort and startup, you will still have to wait for that undo.

-- 
http://www.mladen-gogala.com