AMM and ASMM
Oracle 11gR2 question:
spfile.ora AMM, ASMM configuration
Goal: only use AMM
Issue: both AMM and ASMM are ENABLED !
memory_target 1.5G
memory_max_target 1.28G
sga_target 0
sga_max_size 768M
1. Are AMM and ASMM mutually exclusive.
2. What is expected outcome if both AMM and ASMM are ENABLED
oracletune wrote:
Aman,
Our production system has run since December 2012, with above AMM settings.
The PGA hit ratio was between 98-100 during this period.
The transaction load have been constant and consistent.
Only change: Yesterday re-organized 10million row table [21G] -> to Interval Partitioned Table.
Result a: Fast-Fast query response from Partitioned Table
Result b: New Problem PGA hit ratio 63.52That sounds pretty good as far as problems go. You should look at the plans for your Result a. It could very well be that you've done something like convert a PGA thrashing query to a non-thrashing query, or perhaps even a direct read, leaving only other random queries to hit the PGA. Is anything else slower? Rather than result b being a problem, it could be you just demonstrated the back-asswardness of hit ratios - things may be "better" with a "worse" ratio.
And what sb said.
Similar Messages
-
LOCK_SGA on 11.2 and Linux
hello everyone
is it possible to use LOCK_SGA=true on Oracle Linux 6.2 and UEK (64-bit)?
I am assuming LOCK_SGA=true is required if I use Hugepages on Linux 6 and turn-off AMM and use ASMM
I am trying to get it going and failing.....
using Oracle 11.2.0.3 and Oracle Linux 6.2 with UEK R2, 64-bit
LOCK_SGA=true requires MEMORY_MAX_TARGET=0 and MEMORY_TARGET=0 but even when I specify it all in a PFILE, its non-zero and instance does not start.
I tried SGA_TARGET and SGA_MAX_SIZE and PGA_AGGREGATE_TARGET > 0 and no joy.
I must be doing something wrong - searched Oracle 11.2 manuals - found nothing specific re LOCK_SGA.
Perhaps this requires SPFILE and wont work on PFILE at all ?
Sorry, I dont have MOS, perhaps there is a white-paper on "how to enable LOCK_SGA on 11.2 on Linux Hugepages" ?
thanksthanks.
sounds like I hit a known bug then ?
so let me re-phrase the question:
I want to implement Oracle Linux 64-bit Hugepages on Oracle 11g R2 (11.2.0.3)
What are the steps to do so?
I thought LOCK_SGA=TRUE is a requirement to use Hugepages on 64-bit Oracle Linux.
But it looks like LOCK_SGA is not part of the solution, I believe. Right ?
So..... LOCK_SGA=FALSE.
Now, I turned off AMM, enabled ASMM instead, and enabled Hugepages (which are showing) but they are not being used by Oracle (looking at /proc/meminfo they are all unused)
the
ulimit -l
says "unlimited"
many thanks
Edited by: yurib on Jun 1, 2012 4:15 PM -
Sql statement to determine if amm or ASSM is being used
is there a SQL statement I can use to determine if a DB is using AMM or ASSM. Any examples would be greatly appreciated
As I understand it,
select name, isdefault, display_value from v$parameter where name in ( 'memory_target' , 'sga_target', 'statistics_level') order by name;
If memory_target is set non-zero, you are using Automatic Memory Management See Managing Memory
Else if sga_target is set non-zero and statistics_level is TYPICAL or ALL, you are using Automatic Shared Memory Management See Managing Memory and Configuring and Using Memory
I don't know the Linux memory setup, so I can't comment on the huge pages interaction with AMM and ASSM.
Regards,
David -
Sga_target and shared_pool_size
Hi All,
I am using 11GR2 on linux box. few days back , we hit the ora-4031 and we reported the issue to MOS, they then suggested to bump up the value of shared_pool_size to 3G, which is currently 1G.
Here , we are using sga_target and sga_max_size to 30G and also when I checked the current shared pool size, it's shows around 6G.
Since I am using the Automatic memory memory management along with the shared_pool_size parameter set to 1G. So, Do I really need to bump up the shared_pool_size as per MOS. doesn't this simply
means that , irrespective of shared_pool_size being set , oracle will ignore this value and size the shared pool according to the load in the database.
I would really appreciate , if you could offer your expert opinion here.
SQL> show parameter sga_
NAME TYPE VALUE
sga_max_size big integer 30G
sga_target big integer 30G
SQL> show parameter shared
NAME TYPE VALUE
hi_shared_memory_address integer 0
max_shared_servers integer
shared_memory_address integer 0
shared_pool_reserved_size big integer 338899763
shared_pool_size big integer 1G
shared_server_sessions integer
shared_servers integer 0
SQL> select sum(bytes)/1024/1024 mbytes
from v$sgastat where pool = 'shared pool'; 2
MBYTES
6464
Errors in file /db/prod01/oracle/admin/diag/rdbms/prod01/prod01/trace/prod01_m001_19998.trc:
ORA-04031: unable to allocate 12312 bytes of shared memory ("shared pool","unknown object","KKSSP^281","kglseshtTable")
Fri Aug 16 12:49:19 2013
Process m001 died, see its trace fileIf you are on 11.2(as you have said) , I would recommend that you use AMM only. Also, unsetting the shared pool parameter completely would leave it on the mercy of the AMM and that may go wrong as well. Since you are already hitting 4031, probably the best bet would be to set up a fixed, large sized shared pool as suggested by MOS which, beyond its initial value, would be managed by AMM.
Just my 2 cents
Aman.... -
Linux Huge Pages and SGA Sizing
Hello All -
I have a RAC cluster (2 node), with 132GB of total memory on each box.
I am using AMM and have memory_max_target set at 75GB, and memory_target at 70GB.
I am testing the implementation of huge pages in my lab. My lab cluster has 32GB of total memory, so I set my memory_max_target at 24GB, and memory_target at 22GB.
After running my baseline test, I saw the SGA was automatically set at 12.75GB, and PGA at 9.25GB.
My question is as follows:
I set huge pages to have 24GB of memory locked, so would I set my SGA target at 12.75 and pga_aggregate_target at 9.25GB? Then set sga_max at 14.75GB?
Any advice is much appreciated!
Thanks,
MikeHi Mike,
little strange, that you can configure huge pages and still use AMM. Normally these features are mutually exclusive.
MOS Note 361468.1: HugePages on 64-bit Linux
The AMM and HugePages are not compatible. One needs to disable AMM on 11g to be able to use HugePages. See <<Note 749851.1>> for further information.
Regards
Sebastian -
Constant change in SGA components up-down
hello,
we are developement database oracle 10gR2 on windows plateform
ASMM is on.
i am surprised to know that our smart oracle is doing a lot of shrink and grow on SGA components specially shared pool and db buffer cache
is it ok or some thing wrong?
i checked another development DB and found that it should be infrequent..
do i need to turn of the ASMM?
the out put from a query shown this
WHEN COMPONENT OPER_TYPE INITIAL_SIZE/1024 FINAL_SIZE/1024
May-16:19:53:17 DEFAULT buffer cache STATIC 3031040 3014656
May-16:19:53:17 large pool GROW 0 16384
May-16:19:53:17 DEFAULT buffer cache INITIALIZING 3031040 3014656
May-16:19:53:17 ASM Buffer Cache STATIC 0 0
May-16:19:53:17 DEFAULT 32K buffer cache STATIC 0 0
May-16:19:53:17 DEFAULT 16K buffer cache STATIC 0 0
May-16:19:53:17 DEFAULT 8K buffer cache STATIC 0 0
May-16:19:53:17 DEFAULT 4K buffer cache STATIC 0 0
May-16:19:53:17 DEFAULT 2K buffer cache STATIC 0 0
May-16:19:53:17 RECYCLE buffer cache STATIC 0 0
May-16:19:53:17 KEEP buffer cache STATIC 0 0
May-16:19:53:17 DEFAULT buffer cache STATIC 0 3014656
May-16:19:53:17 streams pool STATIC 0 0
May-16:19:53:17 java pool STATIC 0 16384
May-16:19:53:17 large pool STATIC 0 16384
May-16:19:53:17 shared pool STATIC 0 688128
May-16:20:30:59 shared pool GROW 688128 704512
May-16:20:30:59 DEFAULT buffer cache SHRINK 3014656 2998272
May-17:10:50:40 DEFAULT buffer cache SHRINK 2998272 2981888
May-17:10:50:40 streams pool GROW 0 16384
May-17:10:50:42 DEFAULT buffer cache SHRINK 2981888 2965504
May-17:10:50:42 streams pool GROW 16384 32768
May-17:11:04:38 shared pool GROW 704512 737280
May-17:11:04:38 DEFAULT buffer cache SHRINK 2965504 2932736
May-17:11:05:08 shared pool GROW 737280 770048
May-17:11:05:08 DEFAULT buffer cache SHRINK 2932736 2899968
May-17:11:06:01 DEFAULT buffer cache SHRINK 2899968 2883584
May-17:11:06:01 shared pool GROW 770048 786432
May-17:11:20:38 DEFAULT buffer cache SHRINK 2883584 2850816
May-17:11:20:38 shared pool GROW 786432 819200
May-28:15:24:06 DEFAULT buffer cache SHRINK 2850816 2818048
May-28:15:24:06 shared pool any suggations/advice/help??
thank youI saw this a lot on an highly OLTP system once at 10gR2. My SGA was thrashing wildly between buffer cache and shared pool due to a bug in AMM and it was causing performance impact (cursor S pin on X) for no apparent reason. This was particularly prevalent when I moved from a two to a three node cluster.
If you check out v$sga_resize_ops and see if you see a lot of thrashing (sometimes once every three seconds on one instance for 3 hours, then it'd switch to another instance), then my solution was to hardcode the various settings and disable AMM. However, in your case, you're resizing fairly irregularly - probably due to either an undersized component or workload changes. AMM is meant to do what you're describing - is it causing you any issue or is it something that you've noticed and are curious about? -
Inserting to a new table from a remote connection
I am having difficulties with selecting records from a remote database and inserting them into a current table. The following
statement is what I am using:
insert into sme_lc (WTN, BTN, BTNNAME, PICCODE, PICDATE, CRDDDST)
select a.wtn_cd as WTN,
a.old_btn_cd as BTN,
a.old_billing_nm as BTNName,
'LC LOSS' as PICCode,
a.event_dt as PICDate,
b.crd||b.ddst_cd as CRDDDST
from [email protected] a, [email protected] b
where a.old_btn_cd = b.btn_cd
and a.old_btn_cust_cd = b.cust_cd_suf
and b.crd in ('ASM','AMM')
and a.status_cd = 'L'
and a.status_type_cd = 'A'
and a.region_cd = 'O'
and a.event_dt between to_date('02-04-01','yy-mm-dd') and to_date('02-04-30','yy-mm-dd')
and (a.old_btn_cd, a.old_btn_cust_cd, a.wtn_cd) not in
(select old_btn_cd, old_btn_cust_cd, wtn_cd from [email protected] c
where c.old_btn_cd = a.old_btn_cd and c.old_btn_cust_cd = a.old_btn_cust_cd
and c.wtn_cd = a.wtn_cd
and c.status_cd = 'W' and a.status_cd = 'L' and c.event_dt > a.event_dt);
It returns an error of
and a.event_dt between to_date('02-04-01','yy-mm-dd') and to_date('02-04-30','yy-mm-dd')
ERROR at line 15:
ORA-00920: invalid relational operator
ORA-02063: preceding line from EIA.WORLD
If I remove the insert statement and simply run a select query, it runs fine. Does anyone have any ideas? Your help is much
appreciated!Try the following:
insert into sme_lc
(WTN, BTN, BTNNAME,
PICCODE, PICDATE, CRDDDST)
select a.wtn_cd, a.old_btn_cd , a.old_billing_nm,
'LC LOSS', a.event_dt, b.crd||b.ddst_cd
from [email protected] a, [email protected] b
where a.old_btn_cd = b.btn_cd
and a.old_btn_cust_cd = b.cust_cd_suf
and b.crd in ('ASM','AMM')
and a.status_cd = 'L'
and a.status_type_cd = 'A'
and a.region_cd = 'O'
and (a.event_dt between to_date('02-04-01','yy-mm-dd') and to_date('02-04-30','yy-mm-dd') )
and NOT EXISTS
(select 1
from [email protected] c
where c.old_btn_cd = a.old_btn_cd
and c.old_btn_cust_cd = a.old_btn_cust_cd
and c.wtn_cd = a.wtn_cd
and c.status_cd = 'W'
and c.event_dt > a.event_dt); -
SQL strange behavior after 11g upgrade
Hi all,
After upgrading from 10.2.0.3 to 11.2.0.1 (64-bit RedHat), one of my SQLs behaves very strangely.
These are the facts:
- This is upgrade evaluation phase in 'lab' environment
- First run (after db restart) returns results very quickly (acceptable 0.4 secs)
- Every consecutive run is very slow (3-4 minutes)
- Explain Plan did NOT change between runs;
- Number of 'consistent gets' changed drastically (from few thousands to few millions) ...which would usually lead to inefficient data access plan (index), but again, the explain plan did not change!
- If I flush shared_pool, next run is quick again, and the story repeats ... flushing shared_poll is possible in test system only, it's out of question for prod system of course
- All above happens in 'lab' environment, i.e. no multi-user load yet ... practically 'one-session' system
- I have tried both AMM and ASSM memory management(s) and same problem happens regardless (my old 10g database uses ASSM and I am trying to implement ASM in new 11g db)
- On same server, the SQL in question performs correctly if I run it in 10g environment;
- I have updated my 11g Oracle software/db with the latest CPU;
In addition, I have played with 'result cache' in 11g, and putting 'result_cache' hint in my sql resolved the issue. But that will open a lot of new questions for implementation in real app environment ... so at this moment, I don't want to evaluate result cache ...if I don't have to
I am thinking to try this on another server and eliminate eventual issue with some kind of faulty memory on existing server ... after that - to open ticket with Oracle ...
Any similar experience out there? Any thoughts?
Thanks a lot ...Welcome to the forums !
How to post a tuning request
HOW TO: Post a SQL statement tuning request - template posting
When your query takes too long ...
HTH
Srini -
Application Error in running jar file?What to do?
Hi
I have a code for taking snapshots from a mobile camera.
I have installed the jar file in w810i(sony ericcson).when i run the jar file i get an application error?What to do?
I have tried another program to take a snapshot in this mobile.so device supports but could not understand why the error occurs.
Please help me in thisHi supareno,
I did not try the Sun WTK snapshot app, but I do know that the Semacode MIDlet works on W810i, but that one does not try to set any snapshot format and that does not tell me much because I set formats, use AMMS and other JSR extensions when available on the phone, and so on, programmed in a way that should give graceful degradation on lesser devices (and this works on various devices from different brands). I do not have a physical W810i device so I cannot quickly experiment to find out what the W810i can or cannot handle (e.g., trying only null formats on the W810i?). It is unfortunate that the W810i does not give more informative errors than "Application error" at startup.
Thanks. -
I started lightroom 4.1 and it tells me there is a 4.2 upgrade, click here to download.
Why don't I find out about the upgrade from the usual Adobe update service? Who knows man, who knows.
Anyway, I download it and install 4.2. It leaves 4.1 lying around apparently. I uninstall 4.1.
Now 4.2 is asking me for a serial number. I put in my creative cloud serial number and NOPE.
Meanwhile Adobe Application Manager lists only lightroom 4.1. No 4.2 in the list. Doesn't exist I guess.
Boy was I stupid to think a gazillion dollar a year company like Adobe could actually create and deploy a suite of applications that actually work and communicate with each other. I still can't even update my billing info because "there is an error in the store." Do you guys even know what day it is over there?Have you been able to update your billing info?
Most of the applications are updated using the Updater piece of the Adobe Application Manager (AAM). Lightroom is an exception but as noted will let you know if there is an update when launched.
4.2 is an update to 4.1, not a replacement for 4.1. Have you been able to resolve the problem?
If not then at this point run the Adobe Cleaner Tool which you can get from here http://www.adobe.com/support/contact/cscleanertool.html. Then you will be able to install Lightroom again using AMM, and finally update it to 4.2. -
Buffer_cache ration 50%
Helo,
We have 10.2.0.3 on windows 2003 server and the db is oltp with lots of updates and deletes. The db is AMM and I have set the sga_target to 1.2G but have not set the sga_max_size. When I look at the buffer cache hit rations, it is always less than 50%
can anyone advise what should I do to to improve the performance. ThanksYou can fix the sga_target and sga_max_size with the same value. Also buffer cache hit ratio which is < 50% is it from instance startup or from AWR report? Take a AWR report of 15-30 minutes of your database within the high peak time. Check the soft parse and hard parse also the number of full table scan. High number of FTS and hard parse will make the DB slower. Also check the other wait events and high resource taking SQL.
Regards
Asif Kabir -
Hi All,
I am learning Oracle Memory Tuning and gone through some Performance Tuning Articles. My questions are:
- Since Oracle10g has Automatic Memory Management, once we specified Max SGA and MAX PGA what we need to tune in memory?
- Where can I get tuning examples with case scenarios?
Thanks & Regards,
DanishHi Danish,
Since Oracle10g has Automatic Memory Management, once we specified Max SGA and MAX PGA what we need to tune in memory?My company supports hundreds of databases, and I've noticed that while this "autoimated" memory management is OK for small databases, it may not be right for shops with a large or busy system. This is because AMM is "reactive", and waits until it detects a problem before addressing it. I have an alert here about it, how AMM resise operations can cause performance issues, in some cases:
http://www.dba-oracle.com/t_amm_automatic_memory_resize.htm
Oracle has an insatiable appetite for RAM, but there are diminishing marginal returns as you approach full caching.
Your goal is to give him enough RAM to hit the "sweet spot" of optimal pool sizes.
Oracle has pool advisory utilities that can assist you in this, check them out . . . .
Also, I wrote this note especially for people like you, who use AMM and want to tune RAM:
http://www.oracle-training.cc/t_amm_oracle_instance_tuning.htm
Since you are on 10g, I've collected numerous "gotchas" in 10g that you may find useful:
http://www.dba-oracle.com/t_slow_performance_after_upgrade.htm
Where can I get tuning examples with case scenarios?There are literally thousands of examples on the web. I would search like this, for Oracle RAM (or SGA) tuning examples:
http://www.google.com/search?&q=oracle+ram%7Csga+tuning+example
In sum, the best way to learn how to optimize RAM is to dive-in and get your feet wet!
Hope this helps . . .
Donald K. Burleson
Oracle Press author
Author of "Oracle Tuning: The Definitive Reference"
http://www.rampant-books.com/book_2005_1_awr_proactive_tuning.htm -
MIGO - how check batch classification.
Hi all,
When using the transaction Migo (create), the user can enter a certain characteristic through -> batch -> classification for each position. I have to check that the input code is not already used in another posizion.
I found the Badi CACL_VALUE_CHANGE, but at this point I have only the characteristic of a single position.
Does anybody know the internal table which stores the characteristics and at what point can I go to check them?
I also found these internal tables IZH, AMM and ZM, but at the time of final check, they are not available.
I thank you in advanceHi,
You can get the characteristic values from the standad table AUSP , and table CHVW.
Thanks and Regards,
Narayani -
Hi,
i've 11.2 Database.
I know that using AMM, Oracle manages sga and pga together.
But why shink and grow (sga pools) operation impact only shared and buffer and not also the pga?
An example, I've 5Gb of AMM, looking AWR, Oracle set shared pool to about 1,9Gb, buffer to 1,1Gb (sga target around 3,3Gb) and the results to PGA.
Looking Memory Dynamic Components in AWR, i see that shrink/grow operation uses only shared and buffer pool and not the PGA that is not used.
Why this?
Thanks.Mr.D. wrote:
Hi,
i've 11.2 Database.
I know that using AMM, Oracle manages sga and pga together.
But why shink and grow (sga pools) operation impact only shared and buffer and not also the pga?
An example, I've 5Gb of AMM, looking AWR, Oracle set shared pool to about 1,9Gb, buffer to 1,1Gb (sga target around 3,3Gb) and the results to PGA.
Looking Memory Dynamic Components in AWR, i see that shrink/grow operation uses only shared and buffer pool and not the PGA that is not used.
Why this?
Thanks.What I understood from your question is that you dont see reallocation of PGA as you see in SGA resizable components.
[url http://docs.oracle.com/cd/B28359_01/server.111/b28310/memory003.htm] this describes how the PGA is allocated.
Unless there is pressure on oracle to release the memory from PGA to allocate to SGA or vice versa you will not see the change in memory allocation for PGA -
SGA_MAX_SIZE and AMM
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?
MarkAnother 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.
Maybe you are looking for
-
Hard Drive Replacement - SATA Retail Boxes Marked For PC Not Mac
I accidentally posted this question in the " Questions about USB/FireWire devices and RAM" so I'm re-posting it here. I've checked this forum for messages regarding the upgrade of an iMac G5 hard drive from an 80 GB to a higher capacity drive and the
-
Query Data From Database and set Visibility Accordingly
Hi there, i have a case where i have to set the rendered property of some items on my screen according to values retrieved from an sql statement. When i change a value in a listbox, that value must be given to an sql statement as a parameter (select
-
I have 2 versions of JDK installed on my W7 server (64bit). When I start the SQLDEVELOPER application it comes up with a pop-up saying what I already know : "Java version 1.6.0_11 not supported. The minimum version required is 1.6.0_35" It then promp
-
After installing the Zen V software on my Vista computer, I attempted to run the Media?Explorer program but I get an error message stating?that it'sneeds?needs a "file association". What kind of file is it looking for?Is this a Vista compatiablity is
-
T9 - apostrophe causes capital letters to appear
I'm using a 6111. T9 works fine except using an apostrophe automatically capitalises the word. e.g. "Where he" would become "Where He's" I then have to go back to delete the word to get it back to lower case. It's driving me mad; does anyone know a f