Hi Zhu Chao Have you already looked at the article# 223730.1 ? Rajeev On 8/8/07, Zhu,Chao <zhuchao@xxxxxxxxx> wrote: > > we did test in test environment too, with smaller > sort_area_retaineed_size later . > Looks like it does not help much, unfortunately. > > Anyone know how the detail about PGA_AGGREGATE_TARGET works with MTS? > and any experience with upgrade to 10g, session heap ini large pool > nearly doubled? What the hell is in the session heap? Can we dump the > large pool with detail, and does not hang/crash the production > instance? > > On 8/9/07, Alberto Dell'Era <alberto.dellera@xxxxxxxxx> wrote: > > In fact note 30815.1"Init.ora Parameter SORT_AREA_RETAINED_SIZE > Reference Note" > > says: "If the shared server is used, allocation is to the SGA until > > the value in SORT_AREA_RETAINED_SIZE is reached. The difference > > between SORT_AREA_RETAINED_SIZE and SORT_AREA_SIZE is allocated to the > > PGA." > > > > It's also true that the same note says "The default value as reflected > > in the V$PARAMETER dynamic performance view is 0. > > However, if you do not explicitly set this parameter, > > Oracle actually uses the value of the SORT_AREA_SIZE parameter." > > > > So by default the whole sort area is in the Large Pool; one could change > this > > by explicitly setting SORT_AREA_RETAINED_SIZE. > > > > It seems that 223153.1only considers the default case. > > > > HTH > > Alberto > > > > On 8/8/07, Alexander Fatkulin <afatkulin@xxxxxxxxx> wrote: > > > That's interesting since I always thought that it's > sort_area_retained_size what is allocated from large pool and the rest > (sort_area_size - sort_area_retained_size) is allocated out of the PGA > still. > > > > > > Time for some checks indeed. > > > > > > > > > > > > > To Alex: > > > > sort will use large pool in MTS mode. Please check out metalink > note 223153.1. It is very easy to reproduce/prove. > > > > > > > > > > -- > > Alberto Dell'Era > > "the more you know, the faster you go" > > > > > -- > Regards > Zhu Chao > www.cnoug.org > -- > //www.freelists.org/webpage/oracle-l > > >