Friday, December 23, 2011

Non-standard config parameters for ODA

Examining ODA I noticed some non-standard parameters set by Oracle in init.ora for ODA Database.
These parameters allow us to evaluate Oracle view on the ODA:

_enable_NUMA_support     = FALSE  - disables NUMA support. The database server SUN Fire 4370 is NUMA server by nature. It contains two NUMA nodes. But NUMA-functionality is disabled in /etc/grub.conf with numa=off.

But there is HUGE pages configured in /etc/sysctl.conf :
And Oracle uses them, but use the 16g (we configured medium DB):

Starting ORACLE instance (normal)
****************** Huge Pages Information *****************
Huge Pages memory pool detected (total: 26000 free: 26000)
DFLT Huge Pages allocation successful (allocated: 8193)

Therefore, the best size of SGA for O DA is engineered to 52G !
If you set SGA less than 26000 pages, then you loose memory!

_file_size_increase_increment= 2044M

_disable_interface_checking= TRUE
_lm_rcvr_hang_allow_time = 140
_kill_diagnostics_timeout= 140

_gc_undo_affinity        = FALSE
_gc_policy_time          = 0

 _kgl_cluster_lock_read_mostly= TRUE

And intersting lines in alert.log:

LMD0 started with pid=12, OS id=22987
* Load Monitor used for high load check
* New Low - High Load Threshold Range = [23040 - 30720]

No comments:

Post a Comment

Exadata account locked, pam_tally2 and host_access_control

Exam accounts are set up so that they are blocked for 10 minutes after the first wrong password entered. This brings a lot of inconvenien...