Performance Tuning :: How To Determine System Memory Usage By Oracle Processes
Jun 20, 2012
we have 96GB Memory on the UNIX server and 85% of its usage shows oracle processes I want to determine which Oracle processes are taking most of the memory
SGA is around 36G
SGA_TARGET is 40G
PGA is around 4G
the total of around 40-45 GB of usage is understandable but what other oracle process are chewing up the remaining 30-40 GB on the server is not known
load averages: 7.35, 6.46, 6.15; up 248+11:33:21 12:25:03
2202 processes: 2196 sleeping, 1 zombie, 5 on cpu
CPU states: 83.8% idle, 10.5% user, 5.8% kernel, 0.0% iowait, 0.0% swap
Memory: 96G phys mem, 15G free mem, 128G total swap, 128G free swap
PID USERNAME LWP PRI NICE SIZE RES STATE TIME CPU COMMAND
21720 oracle 258 0 0 40G 40G cpu/48 215:28 2.04% oracle
10709 oracle 1 0 2 1816K 1448K cpu/9 0:02 0.90% res_conf_email_
[code]......
determine if a function is worth pinning in memory? I want to come up with a percentage, implying that if the function is already im memory 80%+ of the time then it is not worth it.
We are using the 11g AMM feature and Memory_Target set to 96GB and total RAM on the Server is 128GB Now the top and free shows up only 200MB memory free on the system.
There are 2 process dbw0 and dbw1 which consumes the top memory and this is 30GB per dbw.
Why is the dbw process taking up so much memory when there is not much load on the database.
I am aware that from 11g, memory_target is sufficient for memeory management between SGA and PGA.
what happens if MEMORY_TARGET set to non-zero and SGA_TARGET set to zero values in a 11g database? Does it enable automatic memory management within the SGA?
We regularly hit by ORA-4031 errors. Also, memory_target advisory (v$memory_target_advice) does not show any advisory information.
for eg: memory_max_target = 500m memory_target = 500m
I have a question regarding memory parameters in oracle database 9.2.0.8, especially sga_max_size and db_cache_size. Database server has 32G of ram. Oracle parameter on server shmmax is set to 16G. Is reasonable to set sga_max_size to the same value, and db_cache_size to 80% of that size?
We are investigating performance of SQL executions on a database server and we suspect I/O on the server is an issue
For example one particular statement accesses one row during execution (index access) and still takes 2.4 seconds out of which it does I/O for 1.9 seconds
which of the following sections in the AWR will give us the correct information about the I/O, it is slow or not?
1) Load Profile Logical reads per second Physical reads per second
2) Top 5 Timed Foreground Events waits / time(s) for events like "db file sequential/scattered read" average wait(ms) for events like "db file sequential/scattered read"
3) Foreground Wait Events db file sequential read db file scattered read
4) Wait Event Histogram %of waits <1ms <2ms Disk file operations I/O db file sequential read db file scattered read
5) Wait Event Histogram Detail (64 msec to 2 sec) Wait Event Histogram Detail (4 sec to 2 min)
6) IOStat by Function summary Buffer Cache Readsreads per sec
its possible to have multiple LGWR processes for a single database.If its possible how does the multiple processes write from redo log buffer to online redo log file.
Customer is sending data from legacy system (Source) with the web service which in turn calls a package lying on Oracle server (Target). Now this package is simply inserting data passed by legacy system into master staging table in Oracle database. When they started this process in Sept 2011 then 4 lack records were inserted into staging table. In Oct 11 it was 0 records Nov 11 it was 2 lack records, Dec 11 it was 1 lack records, in Jan 12 it was 1 lac records, Feb 12 73k records, Mar 12 0 records, Apr 12 52k records.
As we see that number of records inserted in the table got reduced with time.. what should be the starting point here since web service is calling that package on the fly, how can i enable trace for that package? I cannot replicate this is Dev as this process is only working in PROD.
Is there any way to tune the following query using lot of CPU:-select description,time_stamp,user_id from bhi_tracking where description like 'Multilateral:%'The explain plan for this is query is:-
Bhi_tracking is used for reporting purpose and contain millions of records.Generally we keep one year data in this table and delete the remaining.Can I drop the table after taking export and then import it back or can i truncatethe table and then insert the rows into it to enhancethe performance.
We are running Oracle Apps 11i on Windows 2003 R2 Enterprise SP2 (32 bit). After enabling /3GB switch parameter in Boot.ini file, the OS recognizes the 8GB RAM and consumes it upto 5.6GB during non-peak hours but as workload increases in evening when end users generate auto receipts the same memory usage ratio reaches upto 6GB resulting the OS gets hang and don't respond at all. Inspite of having 2GB free memory it doesn't consume it. We have to shut down it by pressing the power off button and restart it
The system configuration is as under: HP ProLiant DL380 G6 Intel Xeon 2.4 GHz 8GB RAM (4 x 2 GB)
After the reboots each system was using around 30 GB of memory... now that it's been up for a week memory is up to 98GB used on each system. None of the systems are swapping.
I came across situation where a Nullable column is not using index for 'order by' clause. I added Not Null condition in the 'where' condition but it wasn't useful. I don't wanted to make composite index with not nullable column or with constant or modify column to 'Not Null'
So I carried out test cases and during which I found that in one case the sql statement does 'fast full scan' for data access but does not use index for 'order by' sorting
here are the steps
Initially I kept the column Nullable
SQL> create sequence s5; Sequence created.
SQL> create table t5 as select s5.nextval id,a.* from dba_objects a where rownum<1001; Table created.
SQL> set pages 100 SQL> select column_name,nullable from user_tab_columns where table_name='T5';
Misses in library cache during parse: 1 Optimizer mode: ALL_ROWS Parsing user id: 5
Rows Row Source Operation ------- --------------------------------------------------- 1000 SORT ORDER BY (cr=16 pr=0 pw=0 time=4771 us) 1000 TABLE ACCESS FULL T5 (cr=16 pr=0 pw=0 time=1157 us)
Elapsed times include waiting on following events: Event waited on Times Max. Wait Total Waited ---------------------------------------- Waited ---------- ------------ SQL*Net message to client 68 0.00 0.00 SQL*Net message from client 68 49.49 49.72 ********************************************************************************
select /*+ index(t i5) */ * from t5 t where id is not null order by id
Misses in library cache during parse: 1 Optimizer mode: ALL_ROWS Parsing user id: 5
Rows Row Source Operation ------- --------------------------------------------------- 1000 TABLE ACCESS BY INDEX ROWID T5 (cr=150 pr=0 pw=0 time=5167 us) 1000 INDEX FULL SCAN I5 (cr=71 pr=0 pw=0 time=3141 us)(object id 4673065)
Elapsed times include waiting on following events: Event waited on Times Max. Wait Total Waited ---------------------------------------- Waited ---------- ------------ SQL*Net message to client 69 0.00 0.00 SQL*Net message from client 69 22.89 28.04
Now I modified the 'id' column to Not Null
SQL> alter table t5 modify id not null;
SQL> set pages 100 SQL> select column_name,nullable from user_tab_columns where table_name='T5';
COLUMN_NAME N ------------------------------ - ID N OWNER Y OBJECT_NAME Y SUBOBJECT_NAME Y OBJECT_ID Y DATA_OBJECT_ID Y OBJECT_TYPE Y CREATED Y LAST_DDL_TIME Y TIMESTAMP Y STATUS Y TEMPORARY Y GENERATED Y SECONDARY Y
Misses in library cache during parse: 1 Optimizer mode: ALL_ROWS Parsing user id: 5
Rows Row Source Operation ------- --------------------------------------------------- 1000 SORT ORDER BY (cr=16 pr=0 pw=0 time=2398 us) 1000 TABLE ACCESS FULL T5 (cr=16 pr=0 pw=0 time=1152 us)
Elapsed times include waiting on following events: Event waited on Times Max. Wait Total Waited ---------------------------------------- Waited ---------- ------------ SQL*Net message to client 68 0.00 0.00 SQL*Net message from client 68 37.74 37.91 ********************************************************************************
Misses in library cache during parse: 1 Optimizer mode: ALL_ROWS Parsing user id: 5
Rows Row Source Operation ------- --------------------------------------------------- 1000 TABLE ACCESS BY INDEX ROWID T5 (cr=150 pr=0 pw=0 time=4166 us) 1000 INDEX FULL SCAN I5 (cr=71 pr=0 pw=0 time=3142 us)(object id 4673065)
Elapsed times include waiting on following events: Event waited on Times Max. Wait Total Waited ---------------------------------------- Waited ---------- ------------ SQL*Net message to client 68 0.00 0.00 SQL*Net message from client 68 8.28 8.45
Misses in library cache during parse: 1 Optimizer mode: ALL_ROWS Parsing user id: 5
Rows Row Source Operation ------- --------------------------------------------------- 1000 SORT ORDER BY (cr=6 pr=0 pw=0 time=1342 us) 1000 INDEX FAST FULL SCAN I5 (cr=6 pr=0 pw=0 time=1093 us)(object id 4673065)
Elapsed times include waiting on following events: Event waited on Times Max. Wait Total Waited ---------------------------------------- Waited ---------- ------------ SQL*Net message to client 68 0.00 0.00 SQL*Net message from client 68 1.88 1.89
Questions are
1) Why adding 'where id is not null wasn't enough for the index to get used in 'order by'? 2) While we got 'fast full scan' why index wasn't used for 'order by' clause? 3) Do we need the indexed column in where clause for being used in 'order by clause' too? 4) Do we need 'order by' clause if we are selecting only the indexed column with sequence generated values?
I am using 11g version, and trying to understand the Memory Allocation.
1) The new feature Memory target parameters are dynamic? is in it? But when I set with the ALTER SYSTEM... it was not changed, then I had to restart the DB.
2) With the new 2 parameters (MEMORY_MAX_TARGET, MEMORY_TARGET), the SGA_MAX_SIZE parameter should be 0, right? But in my case, the SGA_MAX_SIZE value is automatically assigned.
I have two tables with same columns(15 of them), I am trying to find difference between two tables using minus operator and then insert in stage table using below code
Issue is table1 has 50 million records table2 is empty
so when first time when we execute this v_collection1,v_collection2 collection will have 50 million records in it which will go in memory, I think this is not good, because going in memory will eat memory and resources while sorting and other activities ?
After fetching records in collection we are inserting that in stage table and then COMMIT so i think that wont be good because committing 50 million will generate large amount of redo?
below is snippet of my code
DECLARE type lst_collection1 IS TABLE OF table1.col1%type INDEX BY binary_integer; type lst_collection2 IS [code].......
I am trying to run a trace file with TKPROF. It throws an error :
MEMORY FAULT
The size of the trace file is 8MB. I tried with some other file trace files bigger in size (10MB) than the above file, it works fine. I tried seeing any permission right is required on that trace file is required for the tkprof to excute, but it has got the same permissions like the other trace files.
The problem is that it is not showing any error number or any other Error Description other than "MEMORY FAULT".
I have few queries on PGA memory management.Since these queries are based on 2-3 examples not exactly same by nature I am summarising it after my understanding for the same
As I understand many workareas can be allocated to a single sql statement and number and sizes of theses workareas is controlled internally by Oracle when Automatic Memory management (PGA_aggregate_target and workarea_size_policy=Auto are set) Since many sessions share the PGA memory, the amount of memory available to each session may vary and if less amount of memory is available for a session for sorting then TEMP tablespace is used
[1] Can we say paging happens and can be checked at this time?
[2] Is there a difference in handling memory while populating pl/sql tables?
As I have encountered ora-04030 while some our developers were populating pl/sql tables but never encountered this error for sorting, hash joins etc Though I don't remember the width of pl/sql table, I am sure the developer used 'LIMIT' clause during bulk collect and still faced the issue.
With a single session on the server, I noticed that the difference in values displayed issuing 'free' command in linux and output values from sesstat did not match at all while there wasn't any heavy OS process involved during the period. I was expecting 'used' and 'free' values displayed by free command (linux) will change and difference would be approximately equals 'before and after values of session pga memory.
[3] Isn't it expected to match?
[4] Can we say in dedicated server, at any moment of time, the SUM of 'session pga memory' represents all the memory used by Oracle SGA, at that point of time?
select sum(value)/1024/1024 "memory in MB" from v$sesstat where statistic#=20;
During one of the tests I got following output (divide value by 10 for my visibility and avoid formatting)
SQL> select a.name, to_char(b.value/10, '999,999,999') value from v$statname a, v$mystat b where a.statistic# = b.statistic# and a.name like '%ga memory%'; 2 3 4
[code]...
The above query is showing above values even when the pl/sql block execution is completed 30 minutes back
[5] Do we call this as 'memory leak' where memory is not released even while some time has passed since session has done something?Of course I am not checking at OS level as mentioned in question [3] above the values won't match!
Still the output of free command for reference(After the pl/sql block executed)
SQL> select * from v$pgastat; NAMEVALUEUNIT aggregate PGA target parameter 524288000bytes aggregate PGA auto target 456256512bytes global memory bound 26214400bytes total PGA inuse 17328128bytes
[code]...
[6] What could be the significance of negative values of 'session pga memory/max'?
Last We have an OLTP system and in the night we run batch processes in 2-4 sessions
Suppose I have 10 GB RAM and with PGA setting of 3.5 GB Now I want the batch process sessions to use max possible memory during nighttime and toggle the setting back in the morning
[7] With above settings (10 GB RAM and 3.5 GB PGA) how can I divide the memory among 4 sessions?
Shall I set 1) PGA_aggregate_target=0 2)Workarea_size_policy=manual 3) Sort_are_size 4) Hash_area_size
[8] What would be approx values for parameter 3 and 4? will it be straight 3.5 GB/ 4?
NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ lock_sga boolean FALSE pre_page_sga boolean FALSE sga_max_size big integer 3G sga_target big integer 2G
from what I read I beleive this will initially grab 2GB of memory on startup and will grab up to to 3GB of memory total for the SGA. The "total" memory can be allocated to different peices of the SGA when needed but will never exceed 3GB. Is this correct or would these settings infringe on any available memory on a system that is already tight on memory?
Secondly, what happens if both these values are set to the same value?
I have a confusion with MEMORY_TARGET and MEMORY_MAX_TARGET parameter. if i set SGA_TARGET, SGA_MAX_SIZE along with MEMORY_TARGET and MEMORY_MAX_TARGET then how oracle will manage the memory? Because as per my understanding if we set MEM
I am running oracle linux 5.7 on VMware Workstation 9 and trying to install Grid 11gr2, during installation my system freezes and I see out of memory errors. Actually Java installer is eating up too much memory over 1gb. Is there a way ? I can limit memory usage by Java process to a certain extent ?
I installed timesten server and client on different machines in LAN, but with the same user and group - ttadmin:ttgroup When I tried to connect to server with ttisqlcs -connStr "DSN=sampleCS"The output gave me error messages as below connect "DSN=sampleCS";S1000: Failed to retrieve IP address of the system. System error: -2The command failed.Done. The related part in sys.ttconnect.ini is set as below
sys.odbc.ini [ODBC Data Sources]sampleCS=TimesTen 11.2.2 Client Driver [sampleCS]TTC_SERVER=192.168.0.206TTC_SERVER_DSN=nredb_ds
What does it mean by "retrieve IP" ?Does it try to resolve IP by the hostname even when I already gave it the ip address?By the way I can ping 192.168.0.206 and telnet at port 53397 with no problem.
I have installed database in one server. I would like to enable AWR into it. Statistics_level is set to Typical. While running the below script to enable the AWR, its gives error -
SQL> exec dbms_scheduler.enable('GATHER_STATS_JOBS'); BEGIN dbms_scheduler.enable('GATHER_STATS_JOBS'); END;
* ERROR at line 1: ORA-27476: "SYS.GATHER_STATS_JOBS" does not exist ORA-06512: at "SYS.DBMS_ISCHED", line 4343 ORA-06512: at "SYS.DBMS_SCHEDULER", line 2802 ORA-06512: at line 1
I am running an Oracle 10.2.0.3 on Solaris 5.9 OS. Front end appplication is PeopleSoft v8.8.From my AWR report I have found below SQL which needs to be tuned:
Looking to understand the difference between instance tuning and database tuning.
What is the difference between these two tuning exercises? I understand that an instance is memory based structures (logical) where as database consists of physical structures.
However, how does one tune a database the physical structure? Does it have to do with file placements/block sizes etc. Would you agree that a lot of that is taken care by ASM now in 11g? What tools are required/available (third party as well as oracle supplied) for these types of tuning scenarios?
Recently i have been working analyzing Oracle Row Chaining and Migration in the database. Is their any way to track the chaining & migration of rows as part of database health checkup.
Else we have analyze the table for detecting row Chaining and Migration.
We are re-designing our App and we have a critical question, what's the best way (in terms of performance) of using TIFF images (about 20K size) with Oracle.
Currently we have a Windows shared file server and we create the tiff images there under a huge directory structure (like /images/ddddmmyy/aa/bb/001, then /images/ddddmmyy/aa/bb/002, etc, etc). Our database is usually in LINUX version 10, 11 or 12. We create about 200,000 images per day, keep them for 60 days and then remove that structure.
Our Web app (developed with .NET) reads those images just to display them on a Web Session (IE).As you can see, what we are doing now works fine. But network sometimes is an issue and also it's hard to keep synchronization with our DR server, backups, etc.
Are we taking the correct approach? It would be better to have the images in CLOB or BLOBS for better performance? If so, As I mentioned, performance is the KEY FACTOR and the most important item to consider in this design.
