I would check your swapiness value is not set to 60 and set it to 1, and
move to hugepages
So that your main memory can’t be swapped out.
On Wed, 20 Mar 2019 at 07:53, <niall.litchfield@xxxxxxxxx> wrote:
HI Upendra
On Wed, Mar 20, 2019 at 2:51 AM Upendra nerilla <nupendra@xxxxxxxxxxx>
wrote:
datafiles will be cached in the FS cache as well as the database blocks
- disk_asynch_io = true
As I think someone else pointed out if you don't use direct IO then the
being cached in the buffer cache. It's fairly unlikely that this will be as
effective as direct IO, especially on a server with multiple databases on
it. Certainly though, so long as the total size of all the databases is
larger than the filesystem cache you will get the swapping you observe.
/proc/meminfo to see how much memory is being used by them. I suspect you
- We aren't currently using HUGEPAGES, I would like to configure
them, but wanted to get this issue fixed before moving into that direction
You have two very large SGAs. I suggest you run grep PageTables
will be surprised, assuming you have several hundred sessions for each
database - which doesn't look unreasonable to assume with such large SGAs
then I expect tens of GB of PageTable entries. .