Tried to prepare my – Grid Control for an upgrade to   See also .

Updated for example the shared_pool_size  as 128MB as minimum size (see below for other  recommended fixed parameter settings),  although I think it also will work while setting the sga_target. Changed the sga_max_size, shutdown, startup… unfortunately: ora-27102 – out of memory. Database couldn’t be started anymore, not even in nomount (seems logic..).   All the notes are pointing to the kernel-parameter kernel.shmall. But when I recalculate, it should be quite sufficient.

Platform: Suse 9.4, 32-bits. I’m only monitoring  for this customer about something more than hundred objects: 25 databases, some application servers, 13 hosts.

Put more memory in the machine: 8GB: Same problem. 83% of the memory appeared to be diskcache, which should be good. Solution for the moment: startup the database with no sga_max_size. Increased the sga_target with 1GB: no problem. Annoying, as I have to bounce the database when I change sga_target.
For the moment no other solution at hand.

Fixed Initalization Parameters (besides 128MB shared_pool)
job_queue_processes    10
db_block_size 8192
timed_statistics TRUE
open_cursors 300
session_cached_cursors 200
aq_tm_processes  1
compatible <currently installed Oracle Database release> (default)
undo_management AUTO
undo_retention 10800
undo_tablespace <any acceptable name>
processes 150
log_buffer 1048576
statistics_level TYPICAL (Note that this value is specific only to Enterprise Manager 10g Repository Database release and later.)
TEMP space (Tablespace)Foot 1  50 MB (extending to 100 MB)
_b_tree_bitmap_plans false (hidden parameter)