Not many of you know that my day-to-day work focus is Oracle E-Business Suite :). Here goes for an EBS-related blog post…
Recently, several of my team’s clients upgraded to version R12.1.3 and faced the same problem related to OACORE memory settings.
Default configuration
OACORE OC4J containers’s default memory parameters are “-Xmx512M -Xms128M”. This is way too small for most EBS systems (unless you are running a sand box environment). You can use the following command to find the setup in your system:
[[email protected] scripts]$ cd $ADMIN_SCRIPTS_HOME [[email protected] scripts]$ ./adoacorectl.sh status | grep "OC4J:oacore" | awk -F"|" '{print $3}' | while read L ; do ps -fp $L | grep -v PID ; done | awk -F"-X" '{print $2 $3}' mx512M ms128M mx512M ms128M [[email protected] scripts]$
My suggestion
As always, it’s kind of difficult to suggest anything specific. (It depends on the situation.) However, instead of using “-Xmx512M -Xms128M”, I would start with “-Xmx2048M -Xms1024M” set up. To change the memory parameters, edit $CONTEXT_FILE and change the following line:
[[email protected] scripts]$ grep s_oacore_jvm_start_options $CONTEXT_FILE ... -server -verbose:gc -Xmx2048M -Xms1024M ... [[email protected] scripts]$
To finalize the setup run:
cd $ADMIN_SCRIPTS_HOME ./adautocfg.sh
And restart OACORE component:
./adopmnctl.sh stopall ./adopmnctl.sh startall
Side notes
- I do believe that if Java and applications’ code would be bug-free, “s_oacore_nprocs” parameter should be set to 1 for most of the systems.
- However, Oracle instructs us to set an additional OARCORE process for all 200 users.
- In most systems I have worked with, the parameter was set to 2.
- Having multiple OACORE OC4J processes per environment makes it difficult to monitor the health of each container.
- If you know how to make sure that each of the OACORE containers generates a correct response (we typically check login screen using the text browser from our monitoring system), then please help me out and share your solution in the comments section of this blog post.
Emergency fix
Just in case you need to fix an issue quickly, you may want to edit $INST_TOP/ora/10.1.3/opmn/conf/opmn.xml file directly and restart OPMN processes to enable new parameters.
P.S. Don’t forget to edit $CONTEXT_FILE to keep it in sync.
As always, any comments are welcome!
Yury
6 Comments. Leave new
Use reload to enable new parameters instead of adopmnctl.sh stopall/startall
adopmnctl.sh reload
You are running adopmnctl.sh version 120.4.12000000.3
opmnctl: reconfiguring opmn…
opmnctl: opmn reloaded successfully…
Good. This option is less destructive for a running production system.
Hey Jan,
Thanks for sharing. I didn’t know about reload option. #everydaylearning
Kudos,
Yury
Hi,
I want to know how many oacore process are running on machine,
OC4JGroup:default_group | OC4J:forms | 14508 | Alive
OC4JGroup:default_group | OC4J:oacore | 9180 | Alive
HTTP_Server | HTTP_Server | 9020 | Alive
But
prompt JDBC Connection Usage Per JVM Process
prompt ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
select machine, process, count(*) from gv$session
where program like ‘%JDBC%’
group by machine,process
order by count(*) desc
/
I will find the howmany process are connect to JDBC ,I want to know what are they,all are oacore or different.
reload do not change value given by
$ADMIN_SCRIPTS_HOME/adoacorectl.sh status | grep “OC4J:oacore” | awk -F”|” ‘{print $3}’ | while read L ; do ps -fp $L | grep -v PID ; done | awk -F”-X” ‘{print $2 $3}’
Did you try to reload OPMN?