Saturday, November 2, 2013

big numbers reported in KGLH0 and KGLHD in 12c make me grumpy

So a long time back I reported after moving our prod OLTP system ( well old job now actually ) up to 11.2 I was seeing ( from detailed shared pool monitoring ) some big memory allocations rolled into KGLH0 and KGLHD ( see here http://grumpyolddba.blogspot.com/2013/03/missing-pcursor-and-ccursor-in-112.html ).  Not something I was used to seeing in 11.1 and below systems ...

Looks like 12c has same "symptom" ... doing some testing on a 32 gb system with a fairly big shared pool size ... ( fixed size of course ).  This is ASM based using CDB and two PDBs ... don't like ( or understand yet ) what I get reported here ...

select * from v$sgastat where bytes > 100000000 order by con_id asc, 3 desc

So my sqlarea in the PDB that I was pushing workload against ( ok I admit testing a lot of hard parsing ) is way under KGLH0 allocations.

I am working on a presentation for shared pool monitoring ( Hotsos 2014 if it gets accepted / collaborate 2014 maybe ) and well this give me good material ... but dang Oracle what the heck is going on here?

No comments:

Post a Comment