How are you (personally) determining LARGE_POOL_SIZE? Could someone be dynamically changing LARGE_POOL_SIZE? If these questions do not result in an explanation for the differences, then it sounds like a good question for Oracle MetaLink (if you have a Support account).
If you discover an explanation on your own, please advise us.
Having discussed this case with Metalink, it was concluded that the "granule" setting which is determined by SGA size is, in my DB, 16M. Therefore the value for large buffer pool has to be in multiples (or granules) of 16M. This is why I set 10M and it actually becomes 16M.
I'm still puzzled: you said in your most recent post, "This is why I set 10M and it actually becomes 16M." Is this a new value you used...different from your original value of "1048576"? (Are you reading "1048576" as 10MB? Because it's actually 1MB.)
Also, if there is a discrepancy between your init<SID>.ora value and your run value, it could be because your Oracle is ignoring init<SID>.ora values since you are in reality using an "spfile". Can you confirm that you are not using an "spfile"?
Sorry,it was set as 1048576 which is of course 1MB but this rounded up to the nearest "granule" and gave 16M. I since updated it to 10000000 and it still went to 16M.
I am not using spfile and the Metalink guy got me to prove that by running a query or two. We even started the database with pfile = ... and again it was 16M.
So its the fact I can have 0, 16M, 32M, etc. Interesting I tried 32768000 and it came back as 16M, I had to use 33554432.
I can post Metalink's actual explanation if it is of interest?
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.