First of all, you should lower your SGA i would set it to 20Gb for a beginning and put it in memory with lock_sga = true
You should lower the db_cache_size as well (must be contained in sga), 32G is too big regarding to the current activity which show around 620M to be currently sufficient ....
so setting a db_cache_size to 2Gb would already be quite generous
it was lock_sga, not log_sga (that was a typo error that i have corrected)
You should consider setting lock_sga to true instead of the current value false, but make sur you fit the oracle recommendation (see the previous pdf and metalink reference previously posted)
I did a typo error : in the view name of the PGA : it is v$pgastat instead of v$pgastats (no trailing "s")
Code:
select * from v$pgastat;
By the way , you might consider setting your lock_sga to true instead of false since it seems you have enough physical memory to hold your SGA into it (this will force the SGA in RAM and not in swap). But do it carefully (if all the RAM is used by the SGA, there will be less RAM for PGA ...) and make sure you follow the oracle recommendations in the pdf link mentionned in my previous post.
Maybe you should generate an html report on your server using the standard report, @$ORACLE_HOME/rdbms/admin/awrrpt.sql
and selecting 2 snaps containing a timeframe in which you encountered performance problem.(The timeframe should not contain downtime, otherwise the reported stat are not relevant) so you could then have a look at it (have a look at heavy sql statements for example).
Does your server run a single oracle instance or more ?
Check how much RAM is used purely by your operating system.
Do you have some other application or ERP running on that server (SAP ? OAP ? other ?) or is it a pure oracle server?
The only reason your box is still alive is that the SGA is not pinned into memory if it is so huge. If you even try to do that the box will no doubt crash after some uptime as the kernel is pinning memory too and you cannot exceed the overall possible value of pinned memory that is about 85%. With sufficient tuning and sufficient memory, AIX is not going to page computational memory as long as the avm value doesnt exceed 97% of total physical memory. Btw I mentioned it before - your minperm is 10% - even by default it is on AIX 6.1 3% and with your memory consumption on the box you should do the same even for AIX 5.3.
The SGA is OVERsized ...
as well as the db_cache_size which is told not to decrease below 32G whereas it seems 75328 x 8192 ~ 620M are currently enough ...
Last edited by ctsgnb; 11-11-2010 at 09:09 AM..
These 2 Users Gave Thanks to ctsgnb For This Post:
Any chance of seeing the initSID.ora file ?
I agree that say 24 Gb total SGA (25% of memory) would be generous unless this database has a history of ORA errors with "normal" values.
An excessively large db_cache_size can itself cause performance issues on databases with a high rate of data change.
A reason that would justify such a big SGA would be if we would have plenty of users and the server would run in shared server mode (the UGA is then located in SGA instead of the PGA).
But it doesn't seem to be our case.
Please give us the output of
The DBA has made some changes as requested by the expert posters in this thread. I will soon post the output as there is some notable difference in the performance.
However, one question to ZXMaus:
You requested to change Your minfree is too high, set it to 3. Your minperm is 10% - even by default it is on AIX 6.1 3%
So, should I set both MINFREE and MINPERM or just one of them to 3%
Hello,
All the commands on AIX are running very slow.
Below is few stats but I didn't find any issue in cpu or memory reosurces
vmstat
System configuration: lcpu=4 mem=6144MB ent=1.00
kthr memory page faults cpu
----- -----------... (2 Replies)
Hello,
Please help me with a script with which I can check long running processes on the database server and the os is AIX.
Best regards,
Vishal (5 Replies)
Hi Guys,
Need you help in one point!
I am working on one shell script which takes following steps :
1. Taking one query result from oracle database
2. Exporting that result to Xls file
3. Mailing that file to my own mail ID
Now, I want to give a threshold limit to one of the column... (0 Replies)
good morning
what is the better solution to examen a P570 ?
because i use topas and nmon, and the results are totally different !!!
with nmon, i have 80% free cpu, and with nmon, i have 90% of used cpu !!!!!!
i take a shot with an intervall of 10s during 10 mn.
thank you (0 Replies)
Hi,
I am running an oracle db 9.2.0.5.0 on ibm p5 550 aix 5.3 with 10g ram, 10G swap space 3 database instances each SGA about 500Meg.
I am getting the following error in my alert log file from time to time:
skgpspawn failed:category = 27142, depinfo = 11, op = fork, loc = skgpspawn3
... (0 Replies)
hi,
how can i diplay:
- the ammount RAM used /free
- ammount of ram used from a pid or prozess
we have the problem, that malloc is returing a NULL pointer errno = 12 ( not enough space).
but i think there is still ram free.
nmon : shows all memory used ?
Memory Use Physical Virtual... (7 Replies)
11-11-2010
