I would also like to hear more about this. Although, I do not believe the
information message indicates any kind of shortage or fragmentation - I get
the impression that it boils down to "Hey, dude, someone just loaded a
really large object into memory." Period. I find the metalink note to be a
bit ambiguous; it talks about "spending a lot of time in finding free memory
extents during an allocate", but the parameter itself is in units of bytes,
not seconds.
We have resorted to setting this parameter "reasonably high" (what does
that mean, anyway?). From my experience, you can get this message even
if/when your sga is not fragmented, so it is not really an indication of
anything noteworthy. Of course, I could be completely misguided, but until
someone tells me so, I will just go my merry way. =)
On 1/11/07, Niall Litchfield <niall.litchfield@xxxxxxxxx> wrote:
>
> I'm being a bear of very little brain here. I'm reading the note above
> since we are seeing KGL Heap Notifications on a production 10.2 system.
> I'm confused as to whether this is telling me that I should resize the
> shared pool, because the shared pool is either fragmented or short of space,
> or if it is just telling me about large objects being loaded into the shared
> pool and I should reset the warning level to something higher(or turn it
> off).
>
> I reckon I'm using 212 of the 304mb shared pool size that this db has
> and so am tending towards the latter theory, but any insight into this
> behaviour in 10.2 would be welcomed.
>
> --
> Niall Litchfield
> Oracle DBA
> http://www.orawin.info
--
Charles Schultz
