Oracle 9.2.0.6

[Deborah Weatherspoon]

Hi all,
I'm having a strange problem with my Oracle 9.2.0.6 database. It keeps creating a new process about every 10 minutes ????? Nothing is dying even after I bring down the database. I also have processes (PROD) running that do not have a database running for it. Is there a kernel parameter that I need to change ??? We have a L3000 version 11i
serviceguard is also running.

[Ninad_1]

What process is it starting ?
ps -eaf
You say that the processes are still getting created even after bringing down the database. This means it may not be problem with your database - if your database is shutting down cleanly - but some oracle user is proabably running some script which is spawning the processes.
Please check other oracle logins and what processes they have started.
Does this problem start after you bring up the database ? - then check your startup script which may be creating the processes.

As I said 1st you need to tell us what processes you are saying are created every 10 mins.
Also check oracle crontab
crontab -l oracle [ if you are logged in as root ]
crontab -l [ if you are logged in as oracle ]

There may be something configured to run thru oracle crontab every 10 mins ?

Regards,
Ninad

[Deborah Weatherspoon]

User Oracle is starting the processes.. This started happening after we went to 9.2.0.6... PROD is a pkg in serviceguard that should not run unless there is a failover.
racle 23103 22853 1 10:30:07 pts/ta 0:00 grep PPRD
oracle 22948 1 0 10:27:38 ? 0:00 oraclePPRD (LOCAL=NO)
adm3:PPRD>ps -eaf | grep PROD
oracle 19778 1 0 09:37:28 ? 0:00 oraclePROD (LOCAL=NO)
oracle 21622 1 0 10:07:32 ? 0:00 oraclePROD (LOCAL=NO)
oracle 20740 1 0 09:52:30 ? 0:00 oraclePROD (LOCAL=NO)
oracle 21336 1 0 10:02:32 ? 0:00 oraclePROD (LOCAL=NO)
oracle 18981 1 0 09:27:27 ? 0:00 oraclePROD (LOCAL=NO)
oracle 21903 1 0 10:12:33 ? 0:00 oraclePROD (LOCAL=NO)
oracle 22214 1 0 10:17:33 ? 0:00 oraclePROD (LOCAL=NO)
oracle 22577 1 0 10:22:34 ? 0:00 oraclePROD (LOCAL=NO)
oracle 20166 1 0 09:42:29 ? 0:00 oraclePROD (LOCAL=NO)
oracle 20459 1 0 09:47:30 ? 0:00 oraclePROD (LOCAL=NO)
oracle 23119 22853 0 10:30:22 pts/ta 0:00 grep PROD
oracle 19264 1 0 09:32:28 ? 0:00 oraclePROD (LOCAL=NO)
oracle 21026 1 0 09:57:31 ? 0:00 oraclePROD (LOCAL=NO)
oracle 22943 1 0 10:27:35 ? 0:00 oraclePROD (LOCAL=NO)

[HGN]

Hi

When you do a cmviewcl do you see the PROD package up and running if yes and if that is not supposed to be running on the node you
are looking then you need to take it down and bring it backup on the correct node.

The upgrade by itself should not start since it is under a package.

Rgds

HGN

[Ninad_1]

You said that PROD is a package which should start only on failover- failover of what ? Check what is causing the start of this package if its not designed to start in the situation you are in. So whatever is designed is not working.
Next thing the LOCAL=NO processes are the connections to oracle from clients [ may be external , may be internal ]
Ideally the connections should get terminated as part of shutdown of database. Are there new connections getting created even after shutdown ? note the timing of the processes to check if any new connections are getting created after database shutdown. Also check if any database processes are still running -
ps -eaf | grep ora_PROD [ if PROD is your database SID , other wise just grep ora_ - but this will report for all databases running ]

Regards,
Ninad

[Deborah Weatherspoon]

Right now I don't know if it is ORACLE or serviceguard that is causing the problem. The only cron jobs scheduled for ORACLE are the startup and shutdown scripts. PPRD processes are still running after it is stopped and since PROD is not really running on adm3 so I have to kill PPRD and PROD processes manually.

NODE STATUS STATE
adm2 up running

PACKAGE STATUS STATE AUTO_RUN NODE
PROD up running enabled adm2
HOME up running enabled adm2

NODE STATUS STATE
adm3 up running

PACKAGE STATUS STATE AUTO_RUN NODE
PPRD up running enabled adm3
adm3:[/] #

[Volker Borowski · Accepted solution]

Hello,
looking at the timestamps it is exactly every 5 Minutes (if you order them :-) and for sure it is a monitoring tool doing this.

Check the listener log ($ORACLE_HOME/network/log) to find out who is doing these connects and where the guy comes from.

I had this with a script cronned this way

echo alter system switch logfile \; | sqlplus "/ as sysdba"

Which worked for creating the logswitch, and even terminated sqlplus due to end-of-file, but did not close the connection orderly because of missing exit. Did not effect the DB at all, but the comm-slots on the listener did a service table overflow after a couple of month :-(

I suspect something similar here.

Good hunting
Volker

[Deborah Weatherspoon]

Bingo !!! You were right Volker, there was an incorrectly configured monitoring tool running. It had the wrong IP address so that is why I was seeing PROD processes running on the wrong node..

Thanks again

[Deborah Weatherspoon]

Volker was the man this time..... There was a monitoring tool running.