Can start or stop Oracle DB.

[Gino Castoldi_2]

Hi,

HP-UX 11.0 Oracle 8.1.7.3 VPO 6.15 MC/sg 11.09

There were numerous sqlplus calls/processes being made in VPO which slowed the system down
so I decided to stop VPO and Oracle.

However on the primary node Oracle will not start up now, it lists these errors:

08/25/03 12:02:56 ERROR ovoareqsdr (Request Sender)(6932) [dcesec.c:925]: Can't

bind to DCE security service: No currently establish
ed network identity for which context exists (dce / sec). (OpC20-2645)
Database: ORA-00020: maximum number of processes (50) exceeded
(OpC50-15)
Could not connect to database ov_net.
Please look if the database processes are running. (OpC50-2)
Connect to database failed. (OpC40-410)

I can start up the "listener" manually but when I run SVRMGRL> startup
it gets these errors:
ORA 01081 "cannot start already-running ORACLE - shut it down first"

If I try to startup Oracle says it needs to be
running. Note: Our database is not in archive mode.

Any ideas as to how to fix this problem?

10 points to any good answer.
Thank you
Gino

[twang]

To kill all oracle processes first, logon as database owner, issue:
# ps -ef |grep "ora_" |grep -v grep |grep $ORACLE_SUD |awk '{print $2}' |xargs -i kill -9 {}

[twang · Accepted solution]

Sorry for typing error, the command should be:
# ps -ef |grep "ora_" |grep -v grep |grep $ORACLE_SID |awk '{print $2}' |xargs -i kill -9 {}

[Helen French]

This is from rp7410 technical white paper:

ORA-00020: Maximum number of processes (600) exceeded. This error may occur when workload testing is invoked. Make sure to increase the parameter of ???processes??? in ???init.tmp??? at setup time or in ???run_init.pops??? at run time. The original script set the value of 200 for version 8.1.5. It is recommended to set the value to 300 at setup time and 400 at runtime for Oracle version 8.1.6 and 8.1.7.

http://www1.itrc.hp.com/service/cki/cache/200000061824906.pdf

[twang]

I have another throught, the ora-01081 error occurs when oracle sends a request to the DBWR process to obtain an exclusive mount lock to make sure that no other instance is mounting the same database. DBWR did not reply to the request after 60 seconds, so oracle times out and generates this exception. Therefore, first of all, verify that the "ORACLE_HOME" and the "ORACLE_BASE" are set to the correct Oracle Home directory for that SID.

[Gino Castoldi_2]

Hi,

Since we are running on a two node cluster I failed VPO over to the fail-over node and VPO started right up.

We checked shared memory/semaphores on the primary node where Oracle has the issues and VPO won't start and there are shared memory segments and semaphores still allocated.

On the fail-over secondary node there were no segments allocated so that's why VPO started up fine.
We're thinking that shared memory, etc, is still allocated on the primary node and that's why Oracle thinks it's still running.

10 points to any good answer.
Thank you
Gino

[Hein van den Heuvel]

>> ORA 01081 "cannot start already-running ORACLE - shut it down first"

This message can still occur even after you
nuked the oracle processes. It means there is still ipcs context around (the SGA!).
To clear that use: svrmgrl> shutdown abort
If it is still unwilling (and a reboot is not
possible), you can try playing with ipcrm and you want to check for a stray file: $ORACLE_HOME/dbs/lk

hth,
Hein.

[Indira Aramandla]

Hi,

Please try the following:-

1. Edit the database init.ora file
2. Locate and increase the processes parameter to a higher value.
3. Save the init.ora file
4. Stop and restart the database
5. Re-connect to the database instance

To set the value for process use this proc.

SESSIONS
Default value: Derived: 1.1 * PROCESSES + 5

TRANSACTIONS
Default value: Derived (1.1 * SESSIONS)

And remember always to check kernel parameters for process limit (maxuproc/maxusers).

You can add the following to your code.SET TRANSACTION READ ONLY and COMMIT WORK RELEASE statements to avoide maxprocess exceeded.

[twang]

Hein van den Heuvel is correct, when you kill Oracle background processes such as PMON and SMON, you must issue 'ipcs' to ensure all memory segments used by those processes are removed.
To remove a held semaphore after a crash,
# ipcs -sa |grep oracle
s 403021 0xc9d9cb2c --ra-r----- oradb dba oracle dba 204 12:17:36 5:00:38

to release semaphore sets 204,
# ipcrm -s 204