The indications from the flush / bounce pattern are that you have a memory leak
in the library cache / shared pool. The concept is that small fragments of
memory get "lost", and making it harder for the optimizer to find large
contiguous chunks. Flushing the shared pool initially clears enough other
stuff to make sufficient contiguous memory, but eventually there are so many
lost small fragments that flush everything else from the pool still doesn't
allow for a large enough free chunk. If something like this is happening I
think you might see some high version counts for some SQL statements, and high
parse time elapsed, lots of time and activity on the dictionary cache and
shared pool latches.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
@jloracle
________________________________
From: Juan Carlos Reyes Pacheco [jcdrpllist@xxxxxxxxx]
Sent: 22 October 2015 16:38
To: Jonathan Lewis
Cc: ORACLE-L
Subject: solved our "problem with views" flushing shared_pool and cache every
hour, not binding problem
Hello Jonathan, thank you for commenting my mail, and everyone :), I explain
you a problem, because you enjoy them :).
