I did crank it up. Oracle wouldn't come up. That was the problem. Weird! That is
why I came to the dba gurus! I couldn't figure it out!

Thanks!

-----Original Message-----
From: Austin, Steve S [mailto:steve.s.[Email Address Removed] Thursday, May 31, 2001 3:03 PM
To: '[Email Address Removed] LazyDBA.com Discussion
Subject: RE: Shared Memory problem on Solaris 7


Oracle goes through some convoluted processing if the SGA won't fit in a
single shared memory segment. It tries this first, but if this fails, it
tries to bring it up in two other designs with increasing fragmentation of
the SGA into smaller memory segments. Search metalink on this; it's quite
odd.

However, I see -no- reason to keep shmmax below its top value of 4G-1. Why
not just crank this up? This is what's recommended in the install guide,
btw...

Also there are good articles on metalink for semmsl, semmns, semmni, etc.

Fwiw, here's what we use with a clustered implementation where one of the
instances can support up to 1500 processes:

set shmsys:shminfo_shmmax=0xffffffff
set shmsys:shminfo_shmmin=1
set shmsys:shminfo_shmmni=100
set shmsys:shminfo_shmseg=100
set semsys:seminfo_semmns=65535
set semsys:seminfo_semmni=255
set semsys:seminfo_semmsl=2048

hth,
Steve
terms:http://www.lazydba.com/legal.html
Oracle LazyDBA home page