This sounds like a memory leak issue here. Do you see the memory footprint of the dedicated server executing the PL/SQL slowly increasing? A quick google on "oracle 10.2.0.5 memory leak" finds several potentially interesting posts. On Thu, May 17, 2012 at 8:49 PM, Ls Cheng <exriscer@xxxxxxxxx> wrote: > Hi all > My developers have a pretty complex batch process (written entirely in > PL/SQL packages & functions) running in 10.2.0.5 RAC database (Solaris 10), > this process runs in around 12 hours time, we have noticed that the process > gets slower and slower over time. > > This process treats customers information, roughly 10000 customers, during > the first hour it takes around 2 seconds per customer, after 1 hour it > tajes 3, after 2 hours 4 seconds and so on, the funny thing is that if we > abort the process and restart again the elapsed time per customer drops > again to 2 seconds but after 1 hour it starts increasing again. I have add > debug information, after each customer treatment we capture v$sesstat and > v$session_event to see what statistics or wait event is increasing when the > elapsed time goes up, the only statistics which increases over time is "CPU > used by this session", basically the process is burning CPU, the rest of > statistics doesnt not vary. > > It is very puzzling and I cannot find more information to debug, does > anyone know how to tackle this sort of problem... :-? > > > TIA > > PD: I ran dtrace (http://www.brendangregg.com/DTrace/procsystime) in the > last test run but the only call which had sustantial variation between a > fast execution and slow execution is pollsys > > > -- > //www.freelists.org/webpage/oracle-l > > > -- Toon Koppelaars RuleGen BV Toon.Koppelaars@xxxxxxxxxxx www.RuleGen.com TheHelsinkiDeclaration.blogspot.com (co)Author: "Applied Mathematics for Database Professionals" www.rulegen.com/am4dp-backcover-text -- //www.freelists.org/webpage/oracle-l