SGA_MAX_SIZE and AMM

Question

Hi, all,
I've been testing using AMM in our newly-upgraded 11gR2 databases. All databases (Prod and non-Prod) are running 11.2.0.3 on AIX 5.3 (64-bit).
I noticed that despite me explicitly setting the following to 0:
alter system set db_cache_size = 0 scope=spfile;
alter system set log_buffer = 0 scope=spfile;
alter system set pga_aggregate_target = 0 scope=spfile;
alter system set sga_max_size = 0 scope=spfile;
alter system set sga_target = 0 scope=spfile;
alter system set shared_pool_size = 0 scope=spfile;
alter system set shared_pool_reserved_size = 0 scope=spfile;
And allowing Oracle 'full control' how it manages all parts of the memory (SGA and PGA)
-- The 'big' database
alter system set memory_max_target = 24G scope=spfile;
alter system set memory_target = 24G scope=spfile;
-- The 'small' database
alter system set memory_max_target = 3G scope=spfile;
alter system set memory_target = 3G scope=spfile;
I bounce both databases and things seem to come up OK. The only exception is that sga_max_size seems to be set to a specific value:
For the 'big' database, it's 16Gb (2/3rds of the total memory_target)
For the 'small' database, it's 1536Mb (1/2 of the total memory_target)
I don't really understand why this is happening. I've double-checked my SPFILE, I've queried v$spparameter and I've read MOS 443746.1
+"If MEMORY_TARGET is set to a non-zero value:+
+If neither is set, they will be auto-tuned without any minimum or default values. We will have a policy of distributing the total memory set by MEMORY_TARGET parameter in a fixed ratio to the the SGA and PGA during initialization. The policy is to give 60% to the SGA and 40% to the PGA at startup."+
BUT I presume that sga_max_size works in the same way that it did when we weren't using AMM - it's the upper limit of the SGA. Not ideal, to be honest, as I can definitely see times when the SGA needs a LOT of buffer cache in Production (and, therefore, lots of SGA)
I do see this in the alert.logs whenever I restart the system:
Using parameter settings in server-side spfile /opt/oracle/product/11.2.0.3/db/dbs/spfilebigdb.ora
System parameters with non-default values:
processes                = 600
sessions                 = 928
timed_statistics         = TRUE
resource_limit           = TRUE
event                    = "10262 trace name context forever, level 160000"
sga_max_size             = 16G
So it's suggesting that it's querying the SPFILE and determining that sga_max_size should be set. But I've double-double-checked and that isn't the case.
Now, I know I can get around this by explicitly setting sga_max_size at memory_target. But I was wondering if anyone else had seen this before too?
Mark

Answer 1

Another workaround is to remove the memory components entirely from the SPFILE (instead of having them set to 0, don't have them in the file at all). This seems to also set the sga_max_size to the memory_target - which is expected behaviour given the MOS note.