DB startup fail with ORA-00064 Errors with huge sga_target over 40Gig_886312.1

rongshiyuan發表於2014-06-21

Database startup can fail with ORA-00064 Errors with huge sga_target of over 40Gig (Doc ID 886312.1)

In this Document
  Symptoms
  Changes
  Cause
  Solution
  References

Applies to:

Oracle Server - Enterprise Edition - Version: 10.2.0.4 to 11.2.0.3 - Release: 10.2 to 11.2
This problem can occur on any platform.

Symptoms

Attempting to start the database with SGA_TARGET of 100GB (with a total of about 500GB of physical memory available) can fail with:

ORA-00064: object is too large to allocate on this O/S (1,15429280)

Changes

The SGA_TARGET instance parameter has been dynamically increased to a value higher than 40GB, like in:

SQL> alter system set sga_target=100G;
SQL> shutdown
SQL> startup

ORA-00064: object is too large to allocate on this O/S (1,15429280)

Cause

The cause of this problem has been identified in Bug:5051962, which has been closed as not-a-bug. It is caused by the granule size needed for the new SGA size to be different than the one currently in use by the instance at the time the ALTER SYSTEM SET SGA_TARGET command is run.

Solution

This is expected behaviour: the granule size is set based on the size of the SGA. If you set a SGA parameter (such as SGA_TARGET) to a value which exceeds the value of the granule size that was calulcated at instance startup, then you must set the _ksmg_granule_size instance parameter as well to choose a larger granule size. As this parameter is not dynamically changeable, an instance restart is required.

The following table, shows granule sizes for specific SGA sizes and releases of Oracle:

SGA size	10g granule size	11g granule size	11g granule size with fix 8813366 installed
<= 1GB	4MB	4MB	4MB
1GB - 4GB	16MB	16MB	16MB
4GB - 8GB	16MB	64MB	16MB
8GB - 16GB	16MB	64MB	32MB
16GB - 32GB	16MB	256MB	64MB
32GB - 64GB	16MB	256MB	128MB
64GB - 128GB	16MB	512MB	256MB
128GB - 256GB	16MB	512MB	512MB
> 256GB	16MB	512MB	512MB

See also document:947152.1 for details on the granule size and its impact.

The following SQL statement can be used to check the actual granule size in use by the instance:

SQL> select bytes from v$sgainfo where name like 'Granule Size';

So in order to implement the solution for the issue at hand, either:

set SGA_TARGET to a value less than or equal to the previous one, or:
set _ksmg_granule_size to a higher value (in this case the value was set to 32MB). When using a text parameter file, add the following parameter:

_ksmg_granule_size=33554432

When using a server parameter file, issue:

alter system set "_ksmg_granule_size"=33554432 scope=spfile;

In either case, restart the instance for the changes to take effect.