Re: PX Deq Credit: send blkd

From: Martin Klier <usn_at_usn-it.de>
Date: Fri, 29 Aug 2008 20:29:52 +0200
Message-ID: <48B84020.5090002@usn-it.de>


Hi Claudia,

Claudia Zeiler schrieb:
> Do you know how many parallel processes are trying to run? We had this wait event.
> It turned out that we were trying to create indexes as 'parallel'
without specifying the degree.
>Oracle defaulted to 57 processes, and we had a FIGHT on our hands. 27
actually managed to run, and the others
> waited on the PX Deq Credit: send blkd wait event. Maybe you too have more parallllllel ism than you intended.

There are many parallel hints in the statement, I guess about 30 :) The system itself has 114 parallel processes (ora_p*) ATM.

One explanation I'd consider as possible is, that the MERGE itself spawns less p-processes than the sub-selects, so that the select's p-processes (the producers) are too fast for the merge (consumer).

But this is a guess, and it would ge great to see some reliable method to track this issue down :)

Regards
Martin

-- 
Usn's IT Blog for Linux, Oracle, Asterisk
http://www.usn-it.de

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Aug 29 2008 - 13:29:52 CDT

Original text of this message